Keywords:
Abstract: Reinforcement learning where the agent learns a policy to optimize a pre-defined reward by interacting with the environment has undergone rapid development and found applications in many areas such as robot navigation and manipulation. In this talk, we focus on inverse reinforcement learning (IRL) where agent learns an appropriate cost function, leading to desired behaviours, based on demonstration data. We introduce a differential dynamic programming (DDP)-based framework for IRL with both open-loop and closed-loop costs and demonstrate that the closed-loop method performs better than the open-loop one in navigating ground robots and UAVs in unknown and dynamic environments. We further discuss a learning-based dynamic weight adjustment scheme for robots operating in human-dense environments and embodied AI based object navigation.

Keywords: Mechatronics Systems, Robotic and Automation Systems, Rescue Systems
Abstract: Mobile robots that have an ability to run through various environments can help people to avoid dangerous situations, such as exploring a disaster area. In order to adapt to various environments, hybrid mobile robots that combine multiple locomotion and select the locomotion according to the terrain have been developed. However, there are no hybrid mobile robots that combine wheeled, wheeled-legged, and rolling locomotion. In this study, we developed a mobile quad-arm robot equipped with wheeled, wheeled-legged, and rolling locomotion capabilities, and proposed and verified a motion trajectory for the transition from wheeled-legged locomotion to rolling locomotion. The developed robot transits from the wheeled-legged posture to the rolling posture using four arms with different configurations and then rolls. The proposed motion trajectory was verified in the experiment where the robot achieved rolling locomotion. The experiments also verified the wheeled and wheeled-legged locomotion capabilities.

Keywords: Mechatronics Systems, Nonlinear Control, Biological and Physiological Engineering
Abstract: The purpose of this paper is to design a kinematics-based path-following control system for the Lizard-Inspired Single-Actuated robot (LISA). LISA is a new type of robot that mimics the quadrupedal walking morphology of lizards with a four-bar linkage mechanism and can realize both propulsion and turning functions with 1 degree-of-freedom drive. To achieve the objectives, this paper takes three approaches: kinematic formulation, control system design, and numerical simulation. In the formulation of kinematics, we formulate LISA's turning angle, stride length, posture, direction of propulsive, curvature, and position in robot coordinates. In control system design, inverse kinematics is replaced with a residual minimization problem of a quadratic evaluation function, i.e., a nonlinear least-squares problem, which is solved numerically by the Levenberg-Marquard method to obtain control inputs. Numerical simulations verify the effectiveness of the designed control system for circular paths. As a result, it is confirmed that the designed control system is effective for the path-following control of LISA.

Keywords: Mechatronics Systems, Mechanical Systems Control, Human Interfaces
Abstract: Collaborative robots require both controls of position and force to perform tasks accurately while avoiding harm to humans. Periodic/aperiodic separation control adjusts quasi-periodic position and quasi-aperiodic force in a single degree of freedom. However, achieving both high-precision position tracking and soft contact remains challenging. This paper proposes quasi-periodic position and quasi-aperiodic admittance control with feedforward control to improve tracking accuracy. In addition, the periodic/aperiodic separation filter integrated with a linear-phase low-pass filter is used. Frequency analysis and an experiment on a coupled motor system validated the effectiveness of the proposed method in improving the tracking performance of both quasi-periodic position and quasi-aperiodic admittance references. Another experiment using a parallel manipulator demonstrated the practicality of the proposed method under different types of human-robot interaction, including continuous contact and sudden impact.

Keywords: Adaptive and Optimal Control, Nonlinear Control, Modeling, System Identification and Estimation

Keywords: Intelligent Control, Process Control, Adaptive and Optimal Control
Abstract: The Proportional-Integral-Derivative (PID) controller is widely used in industrial plant control systems for its straightforward implementation. However, it requires expert knowledge to tune appropriate PID parameters to achieve desired control performance and demands continuous tuning to adapt to external disturbances or system variations. To address these challenges, we propose a novel parameter optimization method using reinforcement learning (RL) technology. Given the sample inefficiency of RL and the difficulty in reward function design, a novel method using RL with model matching is proposed. Specifically, a reference PID controller is designed to produce target process value, and the RL is designed to optimize PID parameters to minimize the error between the target and actual process values. Simulation results on a time-varying control system demonstrate that the RL method achieves step performance that are nearly identical to reference PID controller. Moreover, it adjusts PID parameters dynamically and shows higher robustness to external disturbances and system variations compared to reference PID controller.

Keywords: Adaptive and Optimal Control, Nonlinear Control
Abstract: In this paper, we propose a reinforcement learning (RL) method using Koopman operator and Kernel Trick for discrete-time nonlinear systems with process and observation noise. This method applies a type of reinforcement learning technique to the linear system realized in a high-dimensional space by using the Koopman operator and the Kernel Trick without system identification. The effectiveness of the proposed method is investigated through a duffing oscillator under process and observation noise.

Keywords: Intelligent Control, Mechanical Systems Control
Abstract: In this paper, we provide an example of using action-constrained soft actor-critic (SAC), a family of deep reinforcement learning algorithms, in control of nonlinear systems. Through simulations and experiments of the swing-up control of a rotary inverted pendulum, we confirm the effect of dealing with action constraints in the learning processes of SAC.

Keywords: Computational Intelligence, Nonlinear Control, Mechatronics Systems
Abstract: Distributed drive electric vehicles (DDEVs) feature redundant actuators, offering enhanced control flexibility. To address the trajectory tracking problem in complicated nonlinear system with strong coupling of actuators, this paper proposes a model-free reinforcement learning-based tracking strategy within High- order Fully Actuated (HOFA) system framework. The proposed method integrates feedforward and feedback control. Specifically, a direct parametric design of HOFA system as the feedforward control law, while the feedback control law is obtained by approximating the value function using a neural network. The neural network weights are updated based on the Bellman error and least squares method, ensuring system stability while approximating the optimal control strategy. Simulation results demonstrate that the proposed control method enables smooth and accurate trajectory tracking, validating its effectiveness in improving vehicle stability and path-following performance.

Keywords: Intelligent Control
Abstract: In this paper, we propose a reinforcement learning-based design method of event-triggered control for a probabilistic Boolean network (PBN). A PBN is well known as one of the mathematical models in gene regulatory networks. In event-triggered control for a PBN, the control input is updated only when the state enters a specific set, which can be obtained by Q-learning. In the proposed method, to handle a large-scale PBN with multiple control inputs, the Q-table is constructed for each control input. The effectiveness of the proposed method is presented by a numerical example.

Keywords: Simulation of Large Systems, Autonomous Decentralized Systems
Abstract: This paper proposes a distributed reinforcement learning algorithm for designing decentralized dynamic output feedback controllers for the large-scale networked systems. This algorithm is scalable for the number of subsystems, making it applicable to large-scale networked systems. The effectiveness of the proposed algorithm is shown through a numerical simulation of a linear thermodynamic model

Keywords: Autonomous Decentralized Systems
Abstract: Quadruped animals achieve remarkable mobility by integrating feedforward and feedback control in their decentralized motor control system. This study focused on understanding how local sensory feedback can coordinate whole-body motions of quadruped locomotion. Our previous work proposed a decoupled oscillator-based control for lizard-like walking. It confirmed that local sensory feedback establishes the coordination of the four legs and body trunk movements without neural communication between oscillators. In this paper, we developed a lizard-like robot and conducted a walking experiment for empirical validation. The resulting gait resembles the actual animal gait and establishes trunk-limb coordination similar to the simulation study.

Keywords: Virtual Reality Systems, Autonomous Decentralized Systems, Robotic and Automation Systems
Abstract: In this study, we focused on a female localization behavior of a male silkmoth with the aim of realizing an engineered odor source localization. The male silkmoth has the ability to detect a sex pheromone (Bombykol) emitted by the female and to localize the female. To analyze this ability, we used a virtual reality (VR) framework designed for an insect. This framework allows the provision of controlled ultisensory stimuli and measures behavioral responses under various environmental conditions. Behavioral data obtained through this framework were analyzed to assess how multisensory input affects adaptive localization. Based on the results of this analysis, we developed a multisensory integration model for odor source localization in complex environments.

Keywords: Robotic and Automation Systems, Intelligent Systems, Intelligent Control
Abstract: Sheepdog system is a control system in which a small number of controller agents (shepherds) indirectly guide and control a large number of agents (flock of sheep). This system is expected to be applied to the guidance and control of biological groups such as birds and schools of fish, and to wide-area exploration and navigation by an operator-teleoperated swarm robot system at disaster sites. Previous studies on the system have used the navigation control based on positional information via wireless communication. However, as the number of robots increases, the amount of communication increases, resulting in delays in system operation. In addition, control methods that use sound or light to estimate the relative positions of robots are prone to malfunctions due to noise.Therefore, this study aims to develop the sheepdog-type group navigation method for a large group of agents. As a first step, this study develops a local sensing-based navigation system fo sheep robots. Specifically, we developed a sheep robot interacting locally by ultrasonic sensors, and conducted navigation experiments using the existing shepherd controller, farthest-agent targeting control. The results of the experimental verification suggest that the control of the shepherd robot is scalable to the number of sheep robots.

Keywords: Robotic and Automation Systems, Intelligent Systems, Intelligent Control
Abstract: This study proposes a navigation strategy for swarm robots to adaptively traverse an unspecified environment. The unspecified environment is an environment that changes unpredictably from moment to moment according to the laws of physics, such as a disaster site or the moon's surface. Such environments are essentially different from previously proposed unknown environments, and cannot in principle be accurately modeled in real-time, i.e., they cannot be known. In contrast, we accept that robots get stuck in an unspecified environment, and we seek to solve this problem through the cooperative navigation of multiple simple swarm robots. In our previous work, we proposed the “BYpassing COmpanions Method for Swarm robot navigation (BYCOMS)” in which each robot moves toward the goal while avoiding surrounding robots. The effectiveness of this navigation method is demonstrated in an environment where there are static areas where robots cannot progress. In this paper, we report that the proposed method is adaptive to changes in the environment where the robot's impassable zone changes dynamically and unpredictably.

Keywords: Network System Integration, Process Automation
Abstract: The Society of Instrument and Control Engineers (SICE) technical committee on industrial networks and system in the industrial application division conducts surveys of industrial control system (ICS) users as part of its activities to help improve ICS cybersecurity. From the user survey results, it has become apparent that while users have concerns about cybersecurity, but measures have not been sufficiently implemented. One of the factors hindering the progress of cybersecurity measures is the concern that such measures may negatively impact operations. To alleviate this concern, The first author has worked on identifying key considerations when selecting cybersecurity measures for ICS, as a method to assess whether the disadvantages of security measures are acceptable. This paper describes the background of this initiative and the extracted considerations.

Keywords: Process Automation, Manufacturing Systems, SCADA and Network Systems
Abstract: Digitalisation and increased connectivity have improved the productivity of industrial control systems (ICS) but have also broadened their cyber-attack surface. Existing guidelines such as NIST SP 800-61 and NISTIR 8428 presume the availability of expert-level CSIRTs, a requirement that can be difficult to satisfy for manufacturing sites facing staff and budget constraints. To bridge this gap, the SICE/JEITA/JEMIMA Joint Security Working Group (Joint Security WG) is developing lightweight incident-response documentation that complements the preventive J-CLICS toolset. The proposed concept emphasises operational continuity, clearly defined roles for plant operators and vendors, and procedures that can be executed by non-experts. A comparative analysis with current standards clarifies the unique requirements of production environments. Scenario-based validation and stakeholder feedback will guide further refinement.

Keywords: Manufacturing Systems, SCADA and Network Systems, Components and Devices
Abstract: Digital twins (DTs) are emerging as key technologies for digital transformation in the manufacturing industry. To apply DTs for improvement of security, Hitachi has been developing a simulation platform, Security Digital Twin (SDT), which can evaluate both security measures and their operational side effects using the 3-layer modeling method. While previous research primarily focuses on the simulation algorithm of SDT, this paper presents a web-based SDT prototype application developed for industrial use. The prototype demonstrates practical features such as model creation, simulation of attack scenarios, and visualization of risks and defense strategies, enabling organizations to plan effective security measures while maintaining business continuity.

Keywords: SCADA and Network Systems, Information Management Systems, Components and Devices
Abstract: The networking of control systems has brought various benefits to the field of control. One such benefit is the ability to monitor and control systems remotely, even when no one is present at the site. However, this has also made control systems a target for cyberattacks, which previously focused on information systems. As a result of cyberattacks, numerous incidents have occurred, including disruption of control and theft of information. Responding to incidents caused by cyberattacks requires not only knowledge of control theory and control systems but also knowledge of information systems and information and communications technology. However, the supply of personnel with all of these skills is not keeping pace with demand. Therefore, we have been exploring an exercise program designed to effectively impart knowledge of information systems and information and communications technology to personnel with expertise in the OT field. This paper focuses on the theoretical component of the exercise program and reports on the results of our exploration.

Keywords: Network System Integration, Safety, Environment and Eco-Systems, Information Management Systems
Abstract: Cyber threats that disrupt the normal operation or cause accidents of critical infrastructure systems and industrial automation control systems have become an urgent issue. The governments around the world are advancing regulations and providing political support for OT (Operational Technology, including automation and control system) cybersecurity in addition to IT cybersecurity. Thailand, Japan, and other Asian countries are dealing to these issues. Also, SICE holds study groups on OT cybersecurity to conduct technical discussions and research. Every year, we have Special Panel Session about OT cybersecurity, and discuss the regulations, standards and technologies related to the topic. In this year, we would discuss the current status of OT cyber security in Thailand, Japan and more. This paper introduces this Special Panel Session.

Keywords: Human-Machine Systems, Analytical Measurement
Abstract: In this study, we measured the rider posture, vehicle operation, and gaze on a straight narrow road, and compared driving time, steering angle, head angle, and gaze movement to analyze differences due to skill level. As a result, riders with longer driving times indicated greater changes in steering and gaze movement. In the future, the relationship between posture and gaze will be further analyzed and utilized to support skill acquisition.

Keywords: Human-Machine Systems, System Engineering, Social Systems
Abstract: In recent years, Japan has experienced rapid population aging, accompanied by an increase in elderly drivers. As a result, the proportion of traffic accidents involving senior citizens has increased, and police activities targeting the elderly have become a high priority. Therefore, in this study, we focus on the accident overview — written in Japanese natural language of approximately 30-200 characters, describing the circumstances leading up to each crash — in the dataset provided by the Toyama Prefectural Police Headquarters. We then extract accident patterns using X-means clustering and the TF-IDF method, visualize these patterns via a classification tree of overviews constructed with the UPGMA method, perform accident factor analysis through association analysis, and develop a severity prediction model for future accidents based on the overviews, thereby supporting police activities.

Keywords: Human-Machine Systems, Factory Automation, Safety, Environment and Eco-Systems
Abstract: Traffic accidents in industrial environments have increased in recent years. One possible measure to increase safety is through pointing-and-calling to raise safety awareness of employees on pedestrian crosswalks. However, there remains many challenges introducing this in organizations, such as the lack of immediate feedback, perceived social awkwardness, and lack of reminders. In this study, we propose that the use of a real-time monitoring and feedback system, which automatically evaluates the pointing-and-calling movement of employees, which can be used to promote this habit, thus improving safety at crosswalks. We performed two experiments evaluating the robustness of the detection systems and the accuracy of the scoring system for the movements.

Keywords: Human Interfaces, Human-Machine Systems
Abstract: One of the main challenges faced by individuals with visual impairments is understanding their surroundings due to the lack of visual information. Recent advancements in artificial intelligence (AI) have enabled applications that can automatically describe environments through voice feedback. However, these language-based systems often struggle to convey information about crowded and dynamic environments, such as train stations or busy intersections, making it difficult for users to track multiple moving elements. To address this, we propose the use of spatial audio, leveraging human sound localization capabilities to complement linguistic descriptions and improve spatial awareness. We developed a prototype system and investigated how two factors—the type of sound source (continuous vs. intermittent) and the distance attenuation model (linear vs. exponential)—affect localisation performance for both stationary and moving objects.

Keywords: Human-Machine Systems, Virtual Reality Systems, Human Interfaces
Abstract: Most assembly work in the manufacturing industry is performed according to standardized work instructions. However, conventional work instructions are frequently challenging to comprehend, particularly during complex tasks or when performed by less-experienced operators, thereby reducing work efficiency and increasing operational errors. Furthermore, many foreign workers, who have been hired in large numbers in recent years due to labor shortages, struggle to accurately comprehend Japanese-language instruction manuals. This research proposes an intuitive and user-friendly work support system based on mixed reality (MR). The silhouette puzzle task serves as an illustrative example of an assembly task. A worker wears a MR device and performs the task while following work instructions displayed in the MR space, thereby improving work efficiency. Experimental results reveal differences in puzzle-solving time depending on the types of instructions: written with diagrams, video, or interactive MR instructions. Additionally, we used the NASA-TLX test to subjectively assess the mental workload associated with the task. The results indicate that the use of MR devices effectively improves work efficiency and reduces mental workload.

Keywords: Computational Intelligence, Human-Machine Systems, System Engineering
Abstract: Reinforcement learning is a machine learning technique. Reinforcement learning requires a long learning time. Transfer learning is a method of shortening this time. However, transfer learning is difficult to apply because the destination and source must be similar. Its application is based on empirical rules. Therefore, we have previously proposed a model that can formulate them. This paper especially studies the relationship between the proposed model and the transition rate. The simulations are based on the classical maze problem.The results of the simulations showed positive correlations between metastatic rates and similarity.

Keywords: Mechanical Measurement
Abstract: The masticatory movements of humans are not clearly understood even in the present day. It is known that the decline in the masticatory function associated with aging causes aspiration and increase the risk of aspiration pneumonia. However, measuring masticatory movements using radiation etc. means may have various effects on the human body. Our laboratory has developed a mastication simulator, which mounted a sensor that measures the pressure acting on the palate. In this research, we examine the change in hardness of the bolus up to the swallowing threshold using this device.

Keywords: Mechanical Measurement
Abstract: ACF is used in the circuit boards and electronic components that make up flat panel displays. Mounting of ACF conducts in two steps: a semi bonding process and a main bonding process, however, the bonding process takes time, which is an obstacle to improving productivity. In this research, we focus on a separator used to remove the release film in the ACF bonding device. By introducing the knife-edge separator, we aim to eliminate the bonding in the semi bonding process and improve productivity. Additionally, the position of the ACF laminated with each separator is evaluated by an image processing program.

Keywords: Robotic and Automation Systems, Medical and Welfare Systems, Identification and Estimation
Abstract: This study proposes a muscle-centered acupressure control strategy that enhances tracking accuracy by localizing the muscle center using multiaxial force sensing. The proposed force-feedback-based control scheme reduces slippage and ensures consistent pressure application by dynamically adjusting the pressing direction. Stepwise tracking adapts to localized muscle deformation, delivering accurate and repeatable force. Continuous tracking integrates real-time force feedback to maintain optimal contact posture, resulting in an increase in vertical reaction force of approximately 2 N. A three-dimensional extension of the system improves spatial precision and further suppresses shear forces. Experimental results validate the effectiveness of this force-sensing-based muscle center localization method in improving the stability and precision of robotic acupressure, demonstrating strong potential for applications in automated physical therapy and rehabilitation.

Keywords: Robotic and Automation Systems, Modeling, System Identification and Estimation, Mechanical Systems Control
Abstract: This study proposes a control strategy for accurate robotic eyebrow drawing on viscoelastic surfaces, emphasizing the estimation and sensing of contact force at the pen tip. To characterize human drawing behavior, line-tracing motions were analyzed on both flat and compliant substrates. A viscoelastic model of a skin-like surface was developed, and a force-based control method was implemented to modify the end-effector trajectory in response to estimated contact force deviations. The system integrates multiaxial force sensing to enable real-time measurement of force components, which is essential for compensating compliance-induced trajectory shifts. Experimental validation confirms that the proposed method significantly reduces drawing errors through contact force–centered sensing and control, highlighting the importance of precise force measurement in robot–surface interaction.

Keywords: Virtual Reality Systems, Robotic and Automation Systems, Factory Automation
Abstract: The advancement of Industry 4.0 has created strong demand for secure, interactive, and real-time visualization platforms capable of accurately reflecting the behavior of physical systems. This paper presents a modular, browser-native 3D graphics engine designed specifically for industrial digital twin applications. Built on WebGL and extended from Three.js, the engine supports cross-browser deployment without plugins and integrates core components including a smart model and animation pipeline, a real-time 3D interaction layer, a customizable 2D UI overlay, and multi-protocol data communication. Real-time connectivity is achieved through native support for WebSocket, MQTT, and OPC-UA, secured via a centralized TLS/SSL Security Manager for encrypted communication and authenticated integration. Assets are created using Blender, leveraging its modeling, rigging, and glTF export capabilities. The engine’s Model Loader and Animation Controller are designed to seamlessly load these assets and bind animations to live sensor data or control commands. Use cases include a robotic arm executing pick-and-place operations and a factory floor simulation with real-time feedback and fault visualization. Performance evaluations show stable frame rates (55–60 FPS on CPU), sub-second model loading, and low-latency communication across tested protocols. These results confirm the engine’s suitability for real-time, secure, and scalable deployment in browser-based industrial applications.

Keywords: System Engineering, Human Interfaces, Components and Devices
Abstract: This paper presents a web-based, cross-platform bootloader framework designed to simplify embedded systems development in both engineering training and rapid product prototyping. The system integrates a lightweight UART-based bootloader for PIC24FJ microcontrollers with a browser-hosted firmware updater developed using WebSerial, TypeScript, and React. It enables zero-install firmware updates directly from Chromium-based browsers such as Chrome, Edge, and Opera, eliminating the need for proprietary drivers, IDEs, or programmer hardware. Key features include automated bootloader entry using DTR/RTS control signals, auto baud rate detection, and LRC-based data integrity verification. Firmware updates are managed via a structured binary protocol and organized through dedicated linker scripts that define isolated memory regions for bootloader and application code. The framework was validated through engineering coursework, training sessions, and controlled lab testing under both clean and noisy electrical conditions. Results demonstrated reliable firmware updates with low retransmission rates and full recovery from all detected transmission errors. With consistent performance across Windows, macOS, and Linux, the framework enables hands-free, cross-platform firmware deployment. This work offers a practical, maintainable solution that bridges the gap between transparent educational tools and real-world embedded development workflows

Keywords: System Engineering, Innovative Systems Approach for Realizing Smarter World
Abstract: This study presents the design and evaluation of a Docker-based training platform, Gambling CTF, developed to simulate realistic cybersecurity scenarios focused on website defacement. The objective was to provide hands-on learning environments for students and IT personnel through modular Capture the Flag (CTF) challenges. The platform comprises four challenge types Hidden Information, Brute Force, SQL Injection, and Cross-Site Scripting each containerized for isolation and scalability. Performance was tested using Apache JMeter with simulated user loads ranging from 100 to 2,500. Results showed that while SQL Injection challenges experienced performance degradation under stress, the platform core (CTFd) maintained full stability. The platform offers a reproducible and secure environment for skill development but is currently limited to known vulnerability types. Future improvements will include automated challenge generation and support for dynamic threat emulation.

Keywords: Intelligent Systems, Adaptive and Optimal Control, Computational Intelligence
Abstract: In the near future, an artificial general intelligence (AGI) that can be carried by each individual is expected to be used to enjoy an epoch-making and fulfilling daily life.

In this study, we envision an action intelligence that can be incorporated into a simple personal AGI that can evolve and be customized autonomously as individuals utilize it. And as a framework for the action intelligence, we focus on observer-based optimal control based on the extended system modeling proposed beforehand in order to generalize the conventional state variable-centered optimal control to that takes into account the situation and environment. Using a simple traffic congestion coping problem as an example, we examine how people take into account situations and environments to plan and control the actions through the extended system model.

Keywords: Intelligent Systems, Robotic and Automation Systems, Autonomous Decentralized Systems
Abstract: This study compares Proximal Policy Optimization (PPO) and Soft Actor-Critic (SAC) deep reinforcement learning algorithms in the 2-vs-2 soccer game environment. We investigate the impact of integrating the Simplicity Bias (SimBa) architecture on learning efficiency and agent performance, evaluated via ELO scores from self-play. While deep reinforcement learning offers powerful tools for complex strategy acquisition, challenges in efficiency and computation persist. Our experiments reveal PPO achieved consistently high performance, whereas SAC struggled, potentially due to hyperparameter sensitivity and replay buffer computational load. SimBa integration accelerated convergence for both algorithms but favored simpler strategies, potentially limiting complex behavior acquisition. These findings provide insights into algorithm selection for multi-agent game AI and the practical trade-offs of architectural biases like SimBa.

Keywords: Social Systems, Entertainment Systems, Identification and Estimation
Abstract: This study investigates fan comments on YouTube related to professional sports. It has been pointed out that professional sports fans have had an “Imagined Community” that talks about sports even before the spread of the Internet. Fans are known to act in response to events such as the relocation or disappearance of their favorite baseball team. Therefore, this study focuses on the “2004 Nippon Professional Baseball realignment” as a discussion topic among Japanese sports fans on YouTube. As for methodology, we collect 5,551 comments from 36 related videos on YouTube, divide them into four categories according to the characteristics of the videos, and analyze the trends of the comments using the machine learning platform DataRobot. In conclusion, we propose future research directions based on the characteristics of the YouTube platform in analyzing social systems from social media.

Keywords: Robust Control
Abstract: In most cases, control systems are subject to uncertainties. Therefore, it is crucial to design robust controllers against such uncertainties. One representative approach for this purpose is gain-scheduled control. Gain-scheduled control is a powerful, robust control method that can be designed when uncertainties that are measurable online, known as scheduling parameters, exist. However, scheduling parameters are not always measurable online, so it is important to relax the conditions of their availability. To address this, a method that estimates the scheduling parameter by a disturbance observer and the recursive least squares method was proposed. Nevertheless, a major drawback of the approach was that the estimate could converge to zero for some numerical examples. In this paper, to overcome this drawback, we propose a novel estimation method that combines a disturbance observer with an adaptive estimation mechanism. This paper also shows the effectiveness of the proposed method through a numerical example.

Keywords: Multivariable Control, Adaptive and Optimal Control
Abstract: This paper proposes a method for guaranteeing a predefined level of optimality for an estimated solution obtained during iteration in the parallel Model Predictive Control (MPC). Primal and dual optimization problems for the parallel MPC are simultaneously solved, and duality theorem in convex optimization theory is used to provide the lower and upper bounds for an optimal value of the cost function in MPC. The proposed method is implemented on low-resource IoT devices, and its effectiveness is validated experimentally.

Keywords: Multivariable Control
Abstract: To analyze and design a system efficiently, it is important to consider in advance what properties of the system can be achieved. Modes of a system are an important algebraic feature in exploring the properties of a system. If mode determination can be achieved by design without going through the process of modeling, then considering modes through mode determination becomes an essential component in formulating design guidelines. This paper gives conditions for data to be mode determinable and design methods directly using the data for this purpose. In particular, by using abstract algebra, we propose a more general form that is independent of the space in which the data is defined.

Keywords: Process Control, Computer Aided Design
Abstract: Set-point tracking and disturbance rejection are important elements in control. Authors have proposed a Predictive Functional Control (PFC) based disturbance rejection controller for measurable disturbances. However, the controller has weakpoints that it cannot handle systems with any complex poles and is not effective for set-point tracking because it is just a feed-forward controller. This paper proposes a PFC-based controller and its design method for addressing set-point tracking and measurable disturbance rejection. Incorporating a state-space model enables control of systems with complex poles. The design uses Virtual Reference Feedback Tuning (VRFT), allowing parameter tuning with one set of operating data without requiring a highly accurate model. Effectiveness was verified by a numerical example. Simulation results showed a larger tracking coincidence point h suppressed input oscillation, albeit with a slight sluggish response. Conversely, a large disturbance coincidence point hd significantly worsened the response. Effective settings for hd require further study. Future work includes considering noise and model mismatch.

Keywords: Multivariable Control, Robust Control
Abstract: This paper proposes a novel data-driven method for stabilization of a linear system with input delay based on the notion of data informativity. Data informativity is a notion that characterizes the necessary and sufficient amount of data for achieving a given control objective, such as stabilization, and was proposed by van Waarde et al. They also extended the previous results to the case of noisy data and derived a condition for the data-driven synthesis of a stabilizing controller. We derive a linear matrix inequality (LMI) condition for the data informativity for quadratic stabilization by considering the structure of the augmented state equation of the input-delay system. The LMI condition enables us to efficiently design the augmented state-feedback controller, which stabilizes every linear input delay system conforming to a given data. The effectiveness of the proposed method is verified through a numerical simulation.

Keywords: Factory Automation, Multivariable Control, Laboratory Automation
Abstract: This paper is concerned with the discretization of a feedback control system capable of preserving the stability of the underlying system for any non-pathological sampling interval. A class of techniques achieving such discretization, called the plant input mapping method, has been developed in our previous studies. An issue remains in the method that it is applicable only to a bi-proper system. The present study is aimed at extending our discretization method to a strictly proper system. An academic example illustrating the features and an experiment on digital implementation are provided.

Keywords: Human Interfaces, Signal and/or Image Processing, Sensors and Transducers
Abstract: This study proposes a machine learning-based gesture classification system using a three-dimensional capacitive proximity sensor unit. To address the nonlinear and variable nature of sensor outputs caused by object direction, distance, and size, three models—Random Forest (RF), Support Vector Classifier (SVC), and Multilayer Perceptron (MLP)—were trained to recognize 19 motion patterns. These patterns include eight directional gestures (up, down, left, right, upper-left, lower-left, upper-right, lower-right), two movement speeds (fast and slow), and three additional actions (left rotation, right rotation, and vertical motion). A total of 17,518 original samples were expanded to 70,064 using data augmentation techniques including scaling, shifting, and noise addition. Among the models, MLP with augmentation achieved the best performance, with a classification accuracy of 99.57% in testing and 89.12% in real-world validation. While classification improved after augmentation, misclassifications due to similar gestures and speed variations remained. Future work will focus on enhancing model generalization and distinguishing adjacent gesture patterns more effectively.

Keywords: Virtual Reality Systems, Agricultural and Bio-Systems
Abstract: In a plant factory, it is necessary to move and transport heavy objects in a harsh environment, such as high temperatures and humidity, while maintaining a sterilized environment and observing the growth of the products. From the viewpoint of cooperation between agriculture and welfare, it is desirable to be able to work with people with physical challenges to support their employment. To date, we have developed a telexistence vehicle for use in plant factories as part of this study. In this study, we will optimize the user interface of the vehicle and evaluate the sense of movement as a telexistece vehicle. Specifically, we developed a system that enables remote control of the vehicle using VR goggles and a smartphone with a touchscreen pad and joystick and evaluated the user interface and the sensation of movement. As future issues, it is necessary to improve VR sickness and to devise a vibrotactile interface that complements the sensation.

Keywords: Agricultural and Bio-Systems, Networked Sensor System, Robotic and Automation Systems
Abstract: Plant factories require precise environmental control to ensure optimal plant growth. However, maintaining uniform conditions throughout a large facility remains a challenge. In this study, we present the development of a mobile monitoring vehicle designed to assess and visualize environmental conditions in plant factories. The vehicle is equipped with a 360-degree camera, environmental sensors (including temperature, humidity, and CO₂), and a QR code reader. QR codes are placed at fixed locations throughout the facility and serve as location markers. When the vehicle reads a QR code, it records the environmental data at that specific point, enabling accurate spatial mapping of the microclimate. The system utilizes an M2M gateway compliant with the IEEE 1888 standard to transmit collected data to a central server in real time. Through a web-based visualization interface, users can monitor environmental fluctuations, detect anomalies, and make informed decisions about environmental control strategies. The use of a mobile platform ensures flexibility and scalability. At the same time, the QR-based localization system provides a cost-effective and reliable method for indoor navigation, eliminating the need for GPS. This monitoring system supports the advancement of smart agriculture by integrating sensing, mobility, and digital visualization technologies into plant factory management.

Keywords: Virtual Reality Systems, Entertainment Systems, Human Interfaces
Abstract: Until now, hands-on demonstrations using the sense of touch have been conducted by connecting paper cups and other objects to devices with vibration sensors and speakers, which lacks freedom in terms of creativity. In this study, we will develop a handmade tactile-auditory fusion media for a more creative workshop. Specifically, we develop a pet that utilizes tactile sensation and squeaks. It does not exist, but it vibrates and squeaks and feels as if it is there.

Keywords: Sensors and Transducers, Biological and Physiological Engineering, Signal and/or Image Processing
Abstract: This study aims to clarify the impact of rising in in-vehicle carbon dioxide (CO₂) concentrations on drivers' cognitive performance and physiological indicators. During summer and winter seasons, vehicle cabins often suffer from inadequate ventilation, leading to elevated CO2 levels due to the confined interior space. Such conditions raise concerns regarding potential adverse effects on human health, particularly in relation to concentration and attention—, which are essential functions for safe driving. We conduct experiments where, participants cognitive performance is assessed under high CO2 conditions after remaining inside a vehicle cabin for 30 minutes without ventilation, Then the same assessment is repeated after cabin ventilation. CO2 concentrations and participants' respiratory rates are monitored using adequate sensors. The results indicate that exposure to a high-CO2 environment leads to a decline in cognitive scores relating to attention and concentration, accompanied by changes in respiratory rates.

Keywords: Opto-Electronic Measurement, Components and Devices
Abstract: This study proposes a surrogate modeling framework for predicting the total luminous flux and chromaticity of white LED packaging using machine learning. Conventional ray-tracing simulations are computationally expensive, particularly for white LEDs containing phosphor, which require complex optical modeling. To address this, we first trained regression models using design parameters—namely, the radiant flux of a blue LED chip and the amount of phosphor—employing multilayer perceptron (MLP), radial basis function network (RBFN), random forest (RF), and support vector machine (SVM). These models achieved high prediction accuracy with coefficients of determination exceeding R²=0.97. In addition, an image-based approach using convolutional neural networks (CNNs) was explored to capture structural features influencing optical performance. Cross-sectional images of white LED packaging were used to train various CNN architectures, including LeNet, ResNet, EfficientNetV2, GhostNetV2, and Vision Transformer (ViT), for comparative analysis. Preliminary results indicate strong correlation between predicted and measured luminous flux values. Feature map analysis revealed that the networks learn to focus on design-relevant regions such as lens curvature, phosphor distribution, and reflectors. These findings suggest that CNN-based models can extract structural insights potentially beyond human intuition, offering a promising approach for data-driven optical design.

Keywords: Human-Machine Systems, Entertainment Systems, Intelligent Systems
Abstract: In recent years, IT technologies have been applied to various sports. As rowing competitions take place on water, there is a distance between the rowers and the coaches during training. Moreover, it was difficult to understand how the oars move in the water during rowing motion. Therefore, we developed the rowing motion visualization system that uses motion sensors to visualize the position of the oar tips, even while the oars are underwater. Furthermore, we developed the visualization system that uses high-precision GNSS and LTE communication to remotely know the boat’s position and speed in real-time. In this study, we extend the rowing motion visualization system and construct the rowing training support system that uses smartphone and cloud server to transmit and visualize not only the boat's position and speed but also rowing motion data in real time. Using this system, we collect data of rowing motions and boat movements of rowers with different skill levels, aiming to clarify the characteristics of rowing motions that contribute most to boat speed. We think this allows us to easily provide feedback to rowers and coaches on the quality of their rowing motions, to improve their skills.

Keywords: Computational Intelligence, Intelligent Systems, System Engineering
Abstract: Deep reinforcement learning is a machine learning method that combines deep learning and reinforcement learning. In deep reinforcement learning, not only the hyper-parameters of deep learning but also the reward functions play important roles to achieve good performance. There are several works in which the hyper-parameters including network structure are optimized using Bayesian optimization, but there have been few works to realize optimizing the reward function in deep reinforcement learning. Therefore, in this study, we attempt to optimize the reward function using Bayesian optimization method. As the value of reward function changes during the Bayesian optimization stage, it cannot be used as the evaluation function for Bayesian optimization. Therefore, we adopt a new objective function to realize the Bayesian optimization of the reward function. In this study, we apply this framework to behavior learning problems in muti-robot environment. We consider the following two types of problems: one is to acquire the behaviors to follow the robot in front of it (following problem) and another is to acquire the behaviors to reach the goal with passing each other (passing problem) in multi-robot environment. We confirm the effectiveness of our framework to optimize the reward function by Bayesian optimization through simulation experiments.

Keywords: Autonomous Decentralized Systems, Computational Intelligence
Abstract: In this study, a consensus algorithm based on fully homomorphic encryption is considered. This approach ensures that the information held by each agent is not disclosed externally, thereby achieving complete privacy preservation. Numerical examples illustrate that the proposed algorithm works well.

Keywords: Virtual Reality Systems, Mechatronics Systems, Components and Devices
Abstract: With recent advances in sensory presentation technology and the widespread adoption of 5G communication, teleoperation and real-time interaction using avatar robots have been realized. However, the structure of existing avatar robots deviates from that of the human body, particularly in terms of visual presentation, causing a sense of dissonance between the images perceived by the operator wearing the head-mounted display (HMD) and the real visual field on the robot side. This dissonance results in diminished immersion and operability. In this study, a visual system for an avatar robot was developed to approximate human visual characteristics. This system mimics human eye movements using eye-tracking data. A fisheye camera mounted on the machine and visual objects on the HMD were rotated dynamically to obtain the original view of the operator. Furthermore, the camera image was projected onto a 220° spherical surface for correction, and a bilateral filter was used to reduce the resolution of the peripheral field of view, thereby reducing the processing load and visual burden. In addition, optical flow was used to compensate for the mis-synchronization of the image and viewpoint that occurs when the camera rotates, resulting in natural and continuous rendering.

Keywords: Components and Devices
Abstract: The risk of cyberattacks is increasing because control systems become more networked. To address this issue from a control engineering perspective, encrypted control has been proposed. Encrypted control ensures secure operation by computing control inputs using encrypted data; however, the encryption process introduces significant computational overhead. Field Programmable Gate Arrays (FPGAs), known for their parallel processing capabilities and low latency, offer a promising solution to this challenge. This study aims to implement a high-speed encrypted controller on an FPGA. Specifically, a PID controller with a disturbance observer was encrypted using the ElGamal cryptosystem with a 143-bit key. The encrypted controller was deployed on an AMD Artix™-7 FPGA (AC701 evaluation kit). The controller’s performance was evaluated by comparing its computation time to that of a CPU implementation.

Keywords: Biological and Physiological Engineering, Sensors and Transducers
Abstract: We developed a portable immunosensing system using smartphone photography and image processing software to determine the concentration of interleukin (IL)-1α in the stratum corneum. A measurable color reaction was first induced by an enzyme-linked immunosorbent assay and captured using a smartphone camera. For imaging, we used a device that guides light from the smartphone's built-in LED and functions as a backlight. Then, we converted the obtained images into RGB values, applied individual coefficients to R, G, and B, and fitted several regression models. We adopted the combination of coefficients and model that produced the highest coefficient of determination as the calibration curve. We also applied this method to stratum corneum samples collected by tape-stripping. For example, when the conventional method measured an IL-1α concentration of 181 pg/mL, our system measured a similar value of 194 pg/mL. Furthermore, when comparing multiple samples, we confirmed a strong overall positive correlation with the results from the conventional method, thereby demonstrating the reliability of our system. In the future, we plan to enhance its applicability for improving quality of life through routine skin assessment by enabling simultaneous acquisition of spectroscopic information in addition to conventional RGB-based image analysis.

Keywords: Temperature Measurement, Safety, Environment and Eco-Systems, Innovative Systems Approach for Realizing Smarter World
Abstract: In Japan, the inspection of exterior wall deterioration is mandatory for large-scale buildings. Drones equipped with infrared cameras have recently become commonly used for this purpose. These inspections rely on detecting surface temperature differences, as deteriorated areas heat up more rapidly than non-deteriorated regions in response to rising ambient air temperatures. As a result, infrared inspections are typically conducted by inspection engineers during morning hours, when temperature gradients are most pronounced. This timing improves the accuracy of deterioration detection. However, since engineers must reach even remote buildings at the optimal time, it has become important to remotely monitor wall surface temperatures in advance to determine the appropriate timing for on-site inspections. In this study, we developed a wall-mounted temperature measurement device that utilizes long-range wireless communication via LTE-M. The system, designed for standalone operation over two weeks, comprises a low-power microcontroller, a GPS module, a temperature sensor, and an LTE-M module. A prototype system fabricated on a printed circuit board transmitted GPS and temperature data to a cloud server. Based on measurements, the device is expected to operate continuously for more than one month.

Keywords: Computational Intelligence, Simulation of Large Systems
Abstract: The cellular automaton model is one of the models employed to simulate traffic flow. In this model, vehicles travel on a road while determining their speed according to a set of rules. The cellular automaton model of traffic flow on a road with multiple lanes incorporates lane change rules to which vehicles change lanes according. In highway contexts, congestion frequently arises in the leftmost lane, which is often obstructed by vehicles directed towards the exit. The implementation of previously proposed lane change rules to emulate this phenomenon shows that vehicles not directed towards the exit do not transit from the congested left lane, despite the availability of the right lane, which is sufficiently spacious for lane changes. The phenomenon under consideration may be attributed to the vehicle's inadequate capacity to locate spaces for lane changes. This study examines the issues with the lane change rules that give rise to this phenomenon and proposes a modified lane change rule to address these issues.

Keywords: Sensors and Transducers, Analytical Measurement, Ubiquitous Healthcare
Abstract: We developed a non-enzymatic uric acid (UA) sensor using molecularly imprinted polymers (MIP) as molecular recognition elements. As interest in personal health monitoring continues to grow, there is an increasing demand for simple and accessible methods to detect biological components such as saliva. MIP are synthetic polymers formed by polymerizing monomers in the presence of a template molecule, which is subsequently removed to create recognition sites that retain the shape and chemical properties of the target. We used a screen-printed carbon electrode (SpCE) as the electrode. Para-aminobenzoic acid (PABA) served as the functional monomer, while UA acted as the template molecule for electropolymerization, thereby facilitating the formation of the MIP layer on working electrode. We employed an Ag/AgCl electrode as the reference electrode. The MIP layer selectively adsorbed UA, and electrons from its oxidation were detected via the p-PABA layer. UA concentrations from 0 to 275 µM were quantified in 25 µM increments based on the resulting potential differences. MIP offer high chemical stability, selectivity, and long-term durability, allowing the construction of enzyme-free, low-cost sensors. The sensor's simple structure facilitates mass production and miniaturization, and it holds promise for real-time UA monitoring via smartphones and applications in wearable health devices.

Keywords: Ubiquitous Healthcare, Analytical Measurement, Biological and Physiological Engineering
Abstract: Here, we present an optical biosensing system designed to measure estradiol, a hormone critical to the female reproductive system. Our system features smartphone-attachment involving custom-designed optics for capturing images of the colorimetric reaction involving 3,3',5,5'-tetramethylbenzidine, which indicates the levels of estradiol immunocomplexes. Additionally, we utilize lab-build image processing software that estimates estradiol concentration through brute-force curve fitting with optimized RGB weighting. The key advantage of our approach lies in its ability to identify the calibration curve models with highest coefficient of determination from multiple function models in a single image processing step. A calibration curve constructed using estradiol standards demonstrated the optimal fit with a four-parameter logistic regression model, yielding a high coefficient of determination (R² = 0.990). Furthermore, a moderate correlation (r = 0.758) was observed between the measurements obtained by this system and those obtained via conventional absorbance spectrophotometry. These findings suggest that estradiol concentration in saliva can be quantified without the use of a microplate reader, indicating the potential of this system as a simple and cost-effective alternative for estradiol quantification. In the future, we aim to automate reagent handling by integrating a microfluidic chip.

Keywords: Signal and/or Image Processing, Embodied-Brain Systems Science
Abstract: This study explores the relationship between auditory stimuli and neural responses by analyzing electroencephalogram (EEG) signals recorded during music listening. A self-built EEG measurement system was developed using Qt, enabling synchronized audio stimulus delivery and real-time EEG acquisition in an eyes-closed resting state. As the auditory stimulus, we selected ”Jupiter, the Bringer of Jollity” from Gustav Holst’s The Planets suite. The recorded EEG signals were processed using a low-pass filter and autocorrelation, followed by slope analysis to quantify temporal trends in the waveform. The results revealed distinct changes in the slope coefficient for the music. At the beginning of the piece, slope values were widely scattered, suggesting neural desynchronization. In contrast, during the refrains and structurally repetitive segments, the slope coefficients converged around zero, indicating increased neural coherence. Toward the end of the piece, slope values tended to shift in the positive direction, possibly reflecting emotional arousal or anticipation. These observations suggest that music induces measurable dynamic transitions in EEG patterns and that slope-based analysis may serve as a sensitive marker for auditory structure processing in the brain. As a future step, we plan to investigate EEG responses during auditory imagery, where participants are asked to imagine the same music after listening. This comparison aims to reveal shared or distinct neural correlate

Keywords: Human-Machine Systems, Human Interfaces
Abstract: This study focuses on human-machine systems in which a human operates a machine. In our previous work, we experimentally discovered that presenting the altered state of the machine to the human can improve the performance of the system. Specifically, the subjects performed a task to balance a cart-type inverted pendulum by applying force to the cart. All the subjects performed better when they were shown the predicted state of the machine rather than its true state. However, the underlying mechanism for this improvement remains unclear. Therefore, this study aims to elucidate this mechanism. We first present a mathematical approach to analyze the effect of feedback of the predicted state on zeros, poles, and gains of 2nd-order Linear Time-Invariant (LTI) systems. We then conduct a reinforcement learning experiment to verify the effect of the predicted state feedback on the operability of the system. Finally, we discuss the mechanism for improving operability by linking the experimental results with the changes in system characteristics caused by the prediction.

Keywords: Modeling, System Identification and Estimation, Computational Intelligence, Process Control
Abstract: This study proposes a Physics-Informed Neural Network (PINN)-based method to model mixing dynamics in stirred tank reactors, which are widely used in batch processes. Scale-up of stirred tanks is critical for increasing production in chemical and materials manufacturing but requires significant time and resources due to changes in internal flow dynamics with tank size. The proposed method estimates the velocity field in stirred tanks by incorporating governing equations and boundary conditions into the learning process. Validation with CFD data showed that the method accurately predicts complex three-dimensional flow fields influenced by stirring blades. Additionally, by introducing design parameters and fluid properties as inputs, the model successfully predicted velocity fields for stirred tanks of varying scales, demonstrating its potential for efficient scale-up in batch polymerization processes.

Keywords: Guidance and Flight Control, Intelligent Control, Robotic and Automation Systems
Abstract: This paper introduces a drone obstacle avoidance method employing Model Predictive Path Integral (MPPI) control integrated with real-time deep learning-based image recognition. Specifically, we utilize the YOLOv8 model to perform real-time obstacle detection and integrate this capability with MPPI control for efficient obstacle avoidance in dynamic conditions. The proposed framework is implemented using Python and validated through experiments conducted with a DJI Tello drone. Experimental results confirm the effectiveness and robustness of the proposed method.

Keywords: Signal and/or Image Processing, Flow Measurement and Control, Modeling, System Identification and Estimation
Abstract: In this paper, we propose a method for accurately detecting water surfaces from low-contrast images, aiming to quantify the water surface dynamics in the water model of a continuous casting mold used in the steel manufacturing industry. Specifically, our method employs a two-stage algorithm: first estimating the initial water-surface positions using Canny edge detection combined with dynamic programming and then refining the surface estimation by removing outliers through mathematical programming, which includes an l1-loss function and a regularization term based on the squared sum of first-order differences. The primary feature of our algorithm is its robustness against outliers by explicitly considering the prior physical information of the smooth temporal and spatial variations of the water surface. Consequently, our method accurately detects water surfaces even in cases where conventional approaches previously misidentified the surface.

Keywords: Modeling, System Identification and Estimation, Signal and/or Image Processing
Abstract: This paper analyzes the Kalman-Takens filter, a model-free, data-driven ensemble Kalman filter based on Takens' embedding theorem. Despite its potential, the filter lacks thorough investigation. To support its understanding and application, the paper derives an upper bound on its estimation error, relating it to noise variance, training data size, and embedding parameters. Numerical experiments with the Lorenz system confirm the theoretical bounds are conservative, offering practical insights into parameter tuning and performance expectations.

Keywords: Computational Intelligence, Identification and Estimation, Nonlinear Control
Abstract: Since the advent of the ``Neural Ordinary Differential Equation (Neural ODE)'' paper, learning ODEs with deep learning has been applied to system identification, time-series forecasting, and related areas. Exploiting the diffeomorphic nature of ODE solution maps, neural ODEs have also enabled their use in generative modeling. Despite the rich potential to incorporate various kinds of physical information, training Neural ODEs remains challenging in practice. This study demonstrates, through the simplest one-dimensional linear model, why training Neural ODEs is difficult. We then propose a new stabilization method and provide an analytical convergence analysis. The insights and techniques presented here serve as a concise tutorial for researchers beginning work on Neural ODEs.

Keywords: Nonlinear Control, Mechanical Systems Control
Abstract: The Jacobian matrix of a vector function is the matrix of all the first partial derivatives of the function. Jacobian matrices are widely used in various applications, and they are useful for stability analysis in control engineering. Jacobian matrices have to be computed numerically in some applications, but their accuracy has not been fully explored. This study proposes an efficient approach for computing numerical Jacobian matrices. The proposed approach presents numerical Jacobian matrices with an arbitrary order of accuracy, whereas a conventional approach presents those with the first order of accuracy. Errors in numerical Jacobian matrices decrease exponentially with the order of accuracy in principle, whereas the computation times increase linearly at most. The proposed approach is useful in a wide range of engineering fields, for example, in the stability analysis of non-closed form systems, such as walking dynamics.

Keywords: Nonlinear Control, Robust Control, Signal and/or Image Processing
Abstract: In this paper, we reveal a novel characteristic in frequency analysis for homogeneous systems. Firstly, we show that every homogeneous control system with a sinusoidal input has an equivalent homogeneous closed-loop system. Particularly, every solution of a homogeneous system for an arbitrary amplitude sinusoidal input can be calculated from the solution under the sinusoidal input with amplitude one. Moreover, the Bode gain plot is mathematically valid for any homogeneous system. Finally, we confirm the effectiveness of the frequency analysis by computer simulation.

Keywords: Adaptive and Optimal Control
Abstract: The background of this research is that clusters of atomic clocks form a network and are controlled to synchronize their time with each other to achieve high-precision time synchronization in the entire network. Highly precision time synchronization enables low latency communication, accurate time stamping, and more efficient use of communication frequencies. In addition to the time synchronization control, an external input is required to broadcast the target synchronization time. The main objective of this study is to optimize the control of this broadcast input.

Keywords: Modeling, System Identification and Estimation
Abstract: Time series modeling plays a crucial role in addressing complex problems across interdisciplinary field, including macro-economic, financial, and engineering. Most of studies primarily concentrate on handle volatility and improve predictive accuracy. The objective of this study is to enhance the time series model, namely, Markov switching autoregressive model with time varying parameter (MS-AR-TVP), to capture complex patterns for prediction. A model error compensator is a key to improve the accuracy of the model by adding the model’s residuals and Kim collapsing procedure to find the model parameter. We integrate MS-AR-TVP model with the model error compensator and name the new model MS-AR-TVP-ME. We apply the proposed model to Thailand GDP dataset from 1993 to 2024 to illustrate the effectiveness of the proposed model. We compare the results of MS-AR-TVP and MS-AR-TVP-ME with that of Long Short-Term Memory (LSTM). MS-AR-TVP-ME outperforms MS-AR-TVP and LSTM in terms of mean square error and mean absolute percentage error.

Keywords: Nonlinear Control, Adaptive and Optimal Control, Computer Aided Design
Abstract: In this paper, we propose a novel design method of finite-time switching signals in nonlinear switched systems described by Lotka-Volterra equations. The problem of designing switching signals is described by combinatorial optimization, for which we propose a relaxation approach to obtain numerically feasible continuous optimization. We show a design example to illustrate the effectiveness of our method.

Keywords: Nonlinear Control, Robust Control, Multivariable Control
Abstract: This paper addresses control design of systems whose state variables describe positive quantities such as concentration, population, and resource. Stationary points of dynamical systems vary with stationary inflow and outflow. A desired stationary point is sometimes unachievable since such flow is unrealizable. Importantly, prohibiting outflow is not necessary for state variables to be positive. The positivity is that a quantity cannot be taken out when zero. Recall that for nonzero stationary points, Metzler matrices and monotonicity with respect to positive orthants, which have been extensively studied in control theory, are irrelevant. To achieve prescribed positive stationary points, this paper proposes a positivizing controller design method based on a single Lyapunov function simultaneously establishing state variable positivity, stability, and disturbance robustness for systems made of cyclic and cascade connections with transportation.

Keywords: Human Interfaces, Biological and Physiological Engineering, Ubiquitous Healthcare
Abstract: Chronic sleep debt is increasingly recognized as a risk factor for depression and various lifestyle-related illnesses. To address this issue, we developed a mobile system that integrates AI-generated voice feedback and gamification principles to promote healthier sleep behaviors. The system features two interaction modes: a multiplayer mode enabling group-based competition via sleep score comparisons, and a single-player mode that rewards users for outperforming their historical averages. Voice-based feedback is delivered using the VoiceVox speech synthesis engine. A two-week experimental study incorporating the Athens Insomnia Scale, reaction time testing, and wearable Fitbit data revealed improvements in both sleep habits and cognitive alertness. These findings suggest that AI-assisted voice rewards and gamified engagement strategies can serve as effective interventions for mitigating sleep debt and enhancing daily performance.

Keywords: Intelligent Systems, Computational Intelligence, Human-Machine Systems
Abstract: The aim of the present study was to clarify the influence of the speaker’s use of pauses on the listener’s perception of humor in episode talks. Episode talks that elicited laughter and those that did not were compared, with a particular focus on the management of pauses. The results revealed a tendency for humorous episode talks to feature longer pauses and shorter speech segments. Additionally, it was observed that in humorous episode talks, rhythm was primarily created through variations in pauses rather than through changes in speech characteristics. Furthermore, regarding the sections preceding the punchline, a general tendency was observed in both humorous and non-humorous episode talks for pause durations to lengthen. The results indicated that humorous episode talks maintained a relatively longer pause-to- speech ratio, suggesting a style of delivery that facilitated easier comprehension of the story. While speech lengthening may contribute to building momentum before the punchline in humorous talks, it is posited that speakers seem to balance this by extending pauses even further, thereby maintaining effective narrative pacing.

Keywords: Relationality-Oriented Systems Design, Human-Machine Systems, Innovative Systems Approach for Realizing Smarter World
Abstract: This study investigates the design of incentive systems centered on the concept of "Care," with a particular emphasis on developing and implementing relationality-oriented systems. Traditional incentive mechanisms often prioritize short-term outcomes and individual motivation, overlooking the significance of interpersonal relationships and long-term social value. In contrast, care-centered incentive systems aim to balance personal motivation with cultivating harmonious social networks, promoting sustainable and holistic system development. This paper first reviews the theoretical foundations of care, incentive mechanisms, and relationality-oriented design. Based on this analysis, it proposes an integrated system framework that combines care principles with incentive strategies. Case studies from various application domains illustrate the framework’s practical value. The findings suggest incorporating care into incentive system design can enhance individual engagement and collective well-being, offering a promising direction for future system innovation.

Keywords: Relationality-Oriented Systems Design, Human-Machine Systems, Innovative Systems Approach for Realizing Smarter World
Abstract: This paper explores and discusses the potential of developing a new type of system characterized by fluid, dynamic, and reconfigurable relationality and assemblage, through the deconstruction of relationality-oriented systems from a care-led perspective. Here, we propose discarding conventional system thinking by introducing a care-led perspective and new directions that are conscious of the relational rather than the individualistic self and the middle voice rather than the active-passive voice view for behavior generation. Furthermore, we propose a new psychological incentive model incorporating motivational mechanisms sensitive to active-passive and middle voice response forms.

Keywords: Human Interfaces, Entertainment Systems, Human-Machine Systems
Abstract: An example of the applications of Image-to-Text AI and/or Speech-to-Text AI is "Automatic Caption Generation" for supporting understanding videos via television (TV) broadcasts or video distribution sites such as YouTube and Netflix. Closed captions not only help the deaf, hearing impaired, and foreign language learners, but also enable all users around the world to access video content in any of various languages. This paper conducts a novel investigation on "What closed captions are appropriate universally or personally for supporting understanding multiple talkers' utterance content?" In addition, this paper conducts a discussion towards implementing an automatic generation system of universal or personalized closed captions with various control variables.

Keywords: Modeling, System Identification and Estimation, Computer Aided Design
Abstract: Accurate prediction of nonlinear system states relies on reliable modelling and parameter estimation techniques. However, direct measurement of system parameters is often infeasible in practice. This study focuses on predicting the short-term behavior of the Lorenz system without direct parameter measurement, using a computer-based approach. A discrete-time model was derived via the continualized discretization method, and system parameters were estimated using the prediction error method by minimizing the root mean square error between observed and predicted states. The resulting model, constructed with the estimated parameters, was then used for short-term state prediction. The results show that the continualized discretization method outperforms the conventional Euler forward difference method in both parameter estimation and prediction accuracy, underscoring the critical role of high-fidelity discretization in modelling nonlinear systems.

Keywords: Modeling, System Identification and Estimation, Identification and Estimation
Abstract: This paper proposes a robust extension of the Sparse Identification of Nonlinear Dynamical Systems (SINDy) framework by incorporating the Huber loss function. The standard SINDy approach, such as STLS and Lasso, employ the L2 norm loss for the error term and are known to be vulnerable to outliers. To address this problem, the Huber loss, which combines the benefits of L2 and L1 norms, has been adopted to ensure robustness against both small errors and large outliers. Since real-world measurement noise is typically Gaussian, the Huber loss is considered well suited for SINDy. To evaluate the proposed method, numerical experiments were carried out on two systems: the Lorenz system without control input, and a two-dimensional drone model with control input. In both cases, noise was synthetically added into training data so that 80% of the data is corrupted with relatively small noise and 20% with larger noise, simulating practical measurement conditions with occasional outliers. The results show that the proposed method achieves higher accuracy in model identification compared to the conventional SINDy under such mixed-noise conditions.

Keywords: Modeling, System Identification and Estimation, System Engineering, Nonlinear Control
Abstract: Diabetes progression is increasingly hypothesized to involve the breakdown of homeostasis—a biological feedback regulation—which may be associated with a bifurcation phenomenon. The dynamical network biomarker (DNB) theory aims to detect qualitative state transitions associated with bifurcation phenomena by identifying DNB nodes whose fluctuations increase near critical points. However, identifying DNB nodes is challenging under high-dimensional and low-sample-size conditions. In this study, we focused on a biomolecular PID feedback control system—referred to as aPID—that exhibits saddle-node bifurcation, assuming that it underlies homeostatic regulation. We analytically show that the aPID controller state inherently satisfies DNB conditions at the bifurcation onset, enabling the a priori identification of DNB nodes without the need for exhaustive data-driven analysis.

Keywords: Modeling, System Identification and Estimation, Process Control
Abstract: In process control, mathematical models are used to represent the characteristics of the target process. However, its characteristics may change over time. Thus, it is desirable to periodically check the accuracy of the model and update it when there is a significant difference between the process and the model. To address this issue, our group has developed a model update method using operational data. The proposed method calculates the posterior distribution of the model parameters, selects suitable input-output data for estimation based on the distribution, and automatically estimates and updates the model using the selected data. However, previous validations have been limited to numerical simulations. Therefore, in this paper, we prove the effectiveness of this method using actual process data.

Keywords: Modeling, System Identification and Estimation, Autonomous Decentralized Systems, Networked Sensor System
Abstract: This paper proposes a secure state estimation method based on virtual distributed reduced-order observers. These observers are constructed by virtually excluding specific sensors or observers from a distributed reduced-order observer system. Unlike conventional approaches that assume undirected or strongly connected graphs, our method targets arbitrary directed graphs composed of multiple independent strongly connected components (iSCCs). By reducing the dimensions of each observer and constructing virtual states for each iSCC, the computational burden is significantly alleviated. We derive necessary and sufficient conditions for secure state estimation and attack detection based on reachability among iSCCs. Finally, we validate the proposed method through numerical simulations, demonstrating its effectiveness in accurately estimating states and detecting attacks even under adversarial conditions.

Keywords: Modeling, System Identification and Estimation, Simulation of Large Systems, Identification and Estimation
Abstract: In this paper, we leverage Koopman mode decomposition to analyze the nonlinear and high-dimensional climate systems acting on the observed data space. The dynamics of atmospheric systems are assumed to be equation-free, with the linear evolution of observables derived from measured historical long-term time-series data snapshots, such as monthly sea surface temperature records, to construct a purely data-driven climate dynamics. In particular, sparsity-promoting dynamic mode decomposition is exploited to extract the dominant spatial and temporal modes, which are among the most significant coherent structures underlying climate variability, enabling a more efficient, interpretable, and low-dimensional representation of the system dynamics. We hope that the combined use of Koopman modes and sparsity-promoting techniques will provide insights into the significant climate modes, thus allowing reduced-order modeling of the climate system and offering a potential framework for predicting and controlling weather and climate variability.

Keywords: Multivariable Control, Nonlinear Control, Human Interfaces
Abstract: This paper presents the possibility of controlling the multi-DoF of soft glove device by constructing angle control in a two-DoF soft glove device composed of thin Mckibben artificial muscles, and demonstrate multi-degree-of-freedom operation.The soft glove device is lightweight and flexible, and therefore has excellent contact with the human body, and is attracting attention as an alternative to hard soft glove devices.Conventional soft glove devices have a low degree of freedom and the direction of the force that can be presented is limited. In addition, posture control has not been constructed, and the possibility of presenting posture information has not been shown.In this paper, we have developed a prototype soft glove device with degrees of freedom in the flexion-extension and adduction-abduction directions using thin Mckibben artificial muscles, and have demonstrated the control performance of the soft glove device using simple posture control.

Keywords: Human Interfaces, Signal and/or Image Processing, Virtual Reality Systems
Abstract: Humans perceive the external world through five senses, all of which are essential for everyday life. As immersive experiences in entertainment and virtual reality become increasingly important, haptic technology has attracted growing interest. While various studies have explored haptic displays that reproduce texture sensations through vibration, their application to compact, widely used devices like smartphones remains limited. This study proposes a smartphone-based system that reproduces the vibrotactile sensation of 2D surface textures by leveraging finger contact metrics, including motion vectors and contact area. Experimental results demonstrate that perceived surface roughness can be modulated more effectively by incorporating contact area into the vibration rendering algorithm.

Keywords: Intelligent Systems, Human-Machine Systems, Robotic and Automation Systems
Abstract: Large Language Models (LLMs) have demonstrated significant capabilities in understanding and generating human language, contributing to more natural interactions with complex systems. However, they face challenges such as ambiguity in user requests processed by LLMs. To address these challenges, this paper introduces and evaluates a multi-agent debate framework designed to enhance detection and resolution capabilities beyond single models. The framework consists of three LLM architectures (Llama3-8B, Gemma2-9B, and Mistral-7B variants) and a dataset with diverse ambiguities. The debate framework markedly enhanced the performance of Llama3-8B and Mistral-7B variants over their individual baselines, with Mistral-7B-led debates achieving a notable 76.7% success rate and proving particularly effective for complex ambiguities and efficient consensus. While acknowledging varying model responses to collaborative strategies, these findings underscore the debate framework's value as a targeted method for augmenting LLM capabilities. This work offers important insights for developing more robust and adaptive language understanding systems by showing how structured debates can lead to improved clarity in interactive systems.

Keywords: Human Interfaces, Virtual Reality Systems, Remote Sensing
Abstract: Although high-speed and tangible are important for realizing interactive 3D displays, it has been difficult to display 3D shapes using only the high-speed projector. In this paper, we focus on anamorphosis, in which a distorted plane image is used to perceive 3D objects, and propose the Dynamic Anamorphosis System. We developed a high-speed feedback system with a delay of less than 6 ms at 1,000 fps and the tangible system that displays 3D objects by anamorphosis using the cylindrical mirror. In addition, combined with face detection, the display is able to correspond to viewpoint changes. We conducted experiments of 3D object displays inside the cylinder and simple physical simulations.

Keywords: Safety, Environment and Eco-Systems, Nonlinear Control, Human-Machine Systems
Abstract: Control barrier functions have attracted significant attention for ensuring the safety of control systems. Recent research has focused on integrating hybrid system frameworks into control barrier function designs to address complex constraints involving multiple state variables and dynamically changing safety constraints. However, most of these studies focused on autonomous systems and ignored the continuity of control inputs. This study addresses hybrid systems with human operator inputs and proposes a control law to ensure safety and input continuity. We confirm the effectiveness of the proposed method through computer simulations.

Keywords: Human Interfaces, Human-Machine Systems, Intelligent Systems
Abstract: This paper explores the perception characteristics of human and machine in interpreting cross-cultural facial expressions. Through a comparative analysis of Thai and Japanese samples, this research investigates the differences in how humans from different cultures and Facial Expression Recognition (FER) systems perceive and interpret cross-cultural facial expressions, identifying the cultural biases present in both. Our study included subjective evaluations from participants in Thailand and Japan, assessing facial expression images from both their own and another culture, followed by a comparison with machine-based assessments. Furthermore, the visual elements surrounding the faces (context) were also included in the subjective evaluations to confirm the influence on recognition capability. The findings reveal that human and machine perception exhibit remarkably similar biases, suggesting that FER systems may mirror the cultural tendencies of human observers.

Keywords: Power Systems Control, Autonomous Decentralized Systems, Discrete Event Systems
Abstract: In recent years, the risk of power outages caused by natural disasters has become a new issue for the power system. As a countermeasure, a mutual aid resilience energy management method using the distributed energy resources of consumers to exchange power for mutual benefit is expected. However, if power distribution lines are damaged during a disaster, there is the vulnerability that the power exchange method cannot be used. Therefore, in this study, we assume power exchange between consumers using Electric vehicles (EVs), rather than distribution lines, as the power exchange method. First, we formulate an optimization problem that expresses power exchange through EV charging and discharging with a mixed integer optimization problem and evaluate a power exchange plan that optimizes the resilience performance of each consumer. Next, consumers are modeled as a multi-agent system and realize power exchange through EV charging and discharging based on Peer-to-Peer (P2P) negotiations in the manner of an event-triggered control method when any consumer’s resilience performance decreases. The two proposed methods are compared with numerical simulations and thus are quantitatively evaluated, confirming the improvement of resilience performance.

Keywords: Adaptive and Optimal Control
Abstract: Abstract - Abstract: Non-linear loads are the ones that are primarily related to the power system network. Harmonic pollution is produced by these loads. Because of the frequency discrepancy with the grid voltage, these harmonic current components do not constitute usable active power. Active power filters are attached to the load in order to lower this harmonic content. The direct control technique and the indirect control method are the two approaches used to minimize errors. Compared to direct current control, indirect current control has a simpler structure and a better harmonic treatment effect. Since just the input variables currents and voltages are monitored, indirect control methods are simpler to implement, whereas direct methods need extremely intricate reference generating processes. This study discusses new, straightforward indirect control ideas for an active power filter (APF) using a PI controller and a Parabolic Current controller. The outstanding simplicity-to-performance ratio attained is the primary benefit over alternative control systems. The idea of virtual impedance emulation serves as the foundation for the suggested control schemes, which aim to provide a system a high-power factor. The SIMULINK/MATLAB environment will be used to simulate the Active Power Filter Control Strategy with Resonant Controller and PI controller utilizing Implicit Closed-Loop Current Control in order to confirm all deductions. The research found that the active power filter

Keywords: Power Systems Control, Adaptive and Optimal Control
Abstract: This paper proposes a day-ahead electricity market model that satisfies a planned balancing policy and maximizes personal profits while minimizing social costs by formulating a DC optimal power flow (DCOPF) model based on an adaptive robust optimization problem. The competitive players participating in this market are modeled as aggregators with multiple energy sources. In particular, the optimization problem includes constraints on transmission capacities. As a result, we determine locational marginal prices (LMPs) that are robust to uncertainties in renewable power generation and contribute to network congestion management. The proposed method is evaluated through numerical simulations of a day-ahead electricity market with five aggregators. The simulation results highlight that each player's renewable energy prediction ability and the introduction of storage batteries are correlated with the increase of personal profit.

Keywords: Power Systems Control, Innovative Systems Approach for Realizing Smarter World, Adaptive and Optimal Control
Abstract: Ensuring stability while achieving efficient market operations in power systems is an increasingly critical challenge, particularly with higher renewable integration. This paper presents an iterative DC OPF approach to integrate a passivity-based stability constraint into the market pricing mechanism, aiming to prioritize stable system behavior through market signals. Simulation results on the IEEE-9 bus system demonstrate that incorporating stability constraint can enhance system stability with only modest reduction in economic efficiency.

Keywords: Power Systems Control, Computational Intelligence, System Engineering
Abstract: This study presents a Reduction of Selective Harmonic Elimination in PWM AC Chopper Using Particle Swarm Optimization technique to improve the quality of the output voltage and minimize harmonic distortion. The proposed method must comply with the technical limitations of the switching angle sequence to achieve the best results. The system is designed to focus on eliminating harmonics in the low order. This directly affects the stability of the electrical system, especially the 3rd,5th,7th,and9th harmonics. Under the setting condition of the output voltage, it should be in the range of 0.1–0.5 p.u.To assess the effectiveness of the technique. The results were compared with the neural network method(ANN) through simulation in MATLAB using R-L loads.It consists of a 50Ω resistor and a 119.3mH inductor. The simulation results show that the proposed method can reduce the 3rd,5th,7th and 9th harmonics more effectively than the ANN method. Better reduces the impact of harmonics on the electrical system. In addition, tests have been conducted on a real test kit to confirm the performance of the system. It was found that the results were likely to be consistent with the simulation results. It shows that the proposed method can be practically applied in electrical systems.Therefore,it can be concluded that the designed system can effectively reduce the harmonic distortion in the electrical system and is a potential choice for the development of techniques to improve the quality

Keywords: Computational Intelligence, Intelligent Systems
Abstract: Many practical multi-objective optimization problems require obtaining robust Pareto solutions to perturbations in decision variables. While some efforts have been made to develop evolutionary algorithms for robust multi-objective optimization, little attention has been paid to designing sound experimental protocols, including appropriate performance indicators, for accurately assessing algorithm performance. To address this gap, this paper conducts a comparative study of performance indicators for robust multi-objective optimization. We compare two popular indicators based on Inverted Generational Distance to investigate which one offers a more reliable assessment of algorithms under the common goal of obtaining robust Pareto solutions. To this end, we employ multi-objective evolutionary algorithms as test algorithms, and evaluate the resulting Pareto solutions using the two performance indicators. Experimental results reveal that performance evaluations vary depending on the chosen indicator, and in some cases, yield results inconsistent with the goal of robust optimization. Based on these findings, we discuss the design of an appropriate experimental protocol, focusing particularly on the choice of performance indicators.

Keywords: Intelligent Systems, Computational Intelligence
Abstract: Accurate trajectory forecasting is critical for the safety and efficiency of autonomous vehicles. Existing models, though accurate, struggle with noise sensitivity, interpretability, and structural complexity, limiting their real-world applicability. This study introduces an LSTM model enhanced with an EMA attention mechanism to address these challenges. The model effectively captures long- and short-term dependencies in vehicle trajectories, emphasizing key temporal features and dynamically adjusting their weights through EMA attention. Unlike traditional mechanisms, EMA reduces computational complexity, mitigates trajectory noise, highlights critical time-steps, and improves model robustness and interpretability. Evaluated on the NGSIM dataset, the EMA-Attention LSTM model demonstrates superior predictive accuracy with low computational overhead, making it suitable for real-time traffic predictions in complex driving environments.

Keywords: System Engineering, Computational Intelligence, Relationality-Oriented Systems Design
Abstract: A risk-averse newsvendor problem is formulated as a chance-constrained problem to maximize profits from the sale of different products. In the newsvendor problem, both of the order quantity and the selling price are optimized for each product. A demand forecasting method based on Bayesian linear regression is used to predict the probability distribution of demand for each product from the selling prices of all products. In order to solve the newsvendor problem, an evolutionary algorithm combined with a mapping-based constraint-handling technique is employed. Finally, the results of numerical experiments show that profits can be increased by optimizing the selling prices for respective products.

Keywords: Intelligent Systems, Computational Intelligence, Robotic and Automation Systems
Abstract: This study presents a predictive approach to improve the decision-making process in autonomous driving applications by modeling and analyzing driver behavior in dynamic environments. Using the CARLA simulator and a random forest-based approach, the proposed method captures interactions between autonomous and human-driven vehicles under various traffic conditions. It evaluates driver behavior through simulated scenarios that include red-light violation, obstacle avoidance, sudden braking, and emergency lane changes. The approach integrates predictive analytics with simulation to enhance adaptability and decision-making accuracy in these complex traffic environments. The findings highlight improvements in safety, situational awareness, and operational reliability, with the random forest-based approach achieving high predictive accuracy across all scenarios. The model effectively identifies critical driving behaviors, with time-to-collision (TTC), speed, and lane clearance emerging as key decision-making factors. Evaluation metrics, including precision-recall and ROC analysis, confirm the approach’s reliability, reinforcing its potential to enhance adaptive and context-aware navigation.

Keywords: Intelligent Systems, Signal and/or Image Processing
Abstract: Iris-based biometric authentication is widely regarded as one of the most secure and reliable authentication methods due to the unique and stable patterns of the human iris. However, traditional authentication systems often suffer from high false positive rates, which may stem from the inclusion of non-informative iris regions in the comparison process. To address this issue, this study explores a thresholding approach that selectively focuses on the most discriminative regions of the iris. The proposed method is evaluated using the CASIA iris image dataset version 2 (CASIA-IrisV2), which includes 60 subjects and a total of 1,440,000 comparison pairs. Experimental results demonstrate a 7.90% increase in recall while maintaining a 3.00% improvement in precision, indicating a significant reduction in false positive without compromising accuracy.

Keywords: Signal and/or Image Processing, Identification and Estimation, Analytical Measurement
Abstract: Contactless respiratory monitoring has emerged as a promising alternative to traditional spirometry, offering advantages in comfort, hygiene, and continuous assessment in both clinical and homecare settings. This study proposes a method for estimating lung volume by tracking chest motion using a 3D Intel RealSense depth camera. By capturing real-time depth data from the thoracic region, the system identifies and tracks key surface landmarks throughout the respiratory cycle. The 3D velocity vector of these landmarks reflects the dynamic expansion and contraction of the chest during breathing. Statistical features such as maximum, minimum, average, and variability of motion are extracted and used to train regression models. Linear regression model based on a three-point chest tracking setup (F3-LR-20) demonstrated the highest predictive performance. The model’s outputs were compared to established reference values from the Baldwin and Japanese Respiratory Society (JSR) equations, with results showing a closer alignment to the JSR standard.

Keywords: Signal and/or Image Processing, Intelligent Systems
Abstract: This paper proposes an enhanced iris recognition system that stores three templates per subject to improve matching accuracy. Instead of relying on a single comparison, each input image is matched against three stored templates, and the final decision is made using one of three voting strategies: majority vote, weighted vote, and biased vote. The biased vote accepts a match if at least one of the three comparisons returns a positive result, based on the observation that the neural network’s match predictions are often correct. Experiments on the CASIA-IrisV2 dataset show that the biased voting strategy achieves the highest recall of 91.25% and accuracy of 95.56%, significantly reducing false negatives. In contrast, the traditional single-template method achieves a lower recall of 73.96% and accuracy of 86.96%. However, using multiple templates increases storage requirements and matching time. Overall, the proposed method offers a more reliable and effective iris recognition system.

Keywords: Signal and/or Image Processing
Abstract: Around 2.2 billion individuals globally experience visual impairments, with roughly 1 billion of those cases being treatable. Analyzing retinal blood vessels is crucial for identifying and distinguishing various retinal vascular conditions, such as diabetic retinopathy, hypertensive retinopathy, corneal opacity, and age-related macular degeneration. Correct diagnosis is essential for successful treatment and to prevent vision impairment. Utilizing machine learning methods to create diagnostic support systems presents a valuable strategy for improving diagnostic precision and minimizing processing duration. This study focuses on enhancing the effectiveness of distinguishing retinal vascular conditions through the application of machine learning techniques. A comprehensive collection of retinal fundus images from patients was gathered and subjected to image processing methods prior to analysis with several machine learning models, such as KNearest Neighbors (KNN), Neural Networks (NN), Adaptive Boosting (AdaBoost), Stochastic Gradient Descent (SGD), and Logistic Regression. The experimental results show that the models created can classify disease groups with an accuracy reaching 97.56%, and they achieve an out-of-sample accuracy of 84.96%. On average, the models achieved an overall classification accuracy of 91.26% across all test conditions. The models showed improved precision and recall when compared to conventional methods.

Keywords: Adaptive and Optimal Control, Multivariable Control, Signal and/or Image Processing
Abstract: Synthetic aperture imaging is a promising concept to achieve unprecedentedly high spatial resolution in satellite-borne optical imagers. A challenge in realizing such an imager consisting of multiple sub-apertures is the development of in-orbit wavefront control methods to correct piston-tip-tilt misalignments between sub-apertures. We investigated the feasibility of image-based alignment approaches using multivariable optimization algorithms. Numerical simulations were conducted to compare two potential algorithmic candidates: stochastic parallel gradient descent (SPGD) and simulated annealing algorithms. Based on these simulation results, we performed laboratory-scale experiments using a tabletop 37-aperture imaging setup and the SPGD algorithm. The demonstration experiments used both static and dynamic scenes as observation targets. These results are discussed to obtain an expected imaging performance of synthetic aperture imagers.

Keywords: Signal and/or Image Processing, Flow Measurement and Control
Abstract: Airflow's visualization, aiming at showing the physical characteristics like shape, density, direction and scope of invisible airflow, is necessary in many areas such as vehicle and interior design. Nowadays, airflow visualization faces a problem: existing visualization methods can be hardly achieved by normal device. Schlieren techniques can visualize some transparent media like airflow. It neither uses extra gas and spray nor affects the observation object. The latest background-oriented schlieren (BOS) method only needs a high-speed camera and a background. Nowadays, event cameras can take the place of high-speed cameras. The event camera outputs lightness changes continuously instead of frames at a fixed frequency. Without the frequency limitation, the event camera can observe tiny and fast changes. To utilize the advantage of event cameras, we propose using event camera to help the normal frame camera gain high-frequency BOS images and thus visualize the frequently changed airflow. The contribution of this paper is as follows: 1. A beam-splitter-based frame and event camera system is implemented and a workflow is proposed to gain aligned synchronous frames and events; 2. The BOS u-Net is proposed to take the BOS frame and event as input to generate a BOS frame of specified time; 3. A BOS frame and event dataset including more than 5,400 BOS frames and corresponding 2,700 events data is made for training and testing the BOS network.

Keywords: Intelligent Control, Adaptive and Optimal Control, Power Systems Control
Abstract: Gear shift design for automotive transmissions to suppress shift shock requires advanced techniques to handle nonlinear behaviors, such as variations in friction coefficients due to changes in relative clutch plate speed and oil temperature. Although reinforcement learning-based control for reducing shift shock has been studied, research focusing on environments where clutch plate friction coefficients change due to mechanical aging remains insufficient. Because conventional reinforcement learning assumes constant state transitions and rewards, it becomes unstable when environmental changes occur, making it difficult to consistently suppress shift shock caused by friction coefficient variations. To address this issue, we propose a reinforcement learning algorithm specifically designed to suppress shift shock. This method varies the discount factor to stabilize learning whenever instability arises. The loss function value is treated as an indicator of learning progress and used as the control variable in a PI-D-based adjustment of the discount factor. In simulation, the proposed approach successfully suppresses shift shock while maintaining learning stability, outperforming conventional methods.

Keywords: Nonlinear Control, Transportation Systems, Intelligent Control
Abstract: This paper tackles the challenge of delivering electric power to evacuation shelters during disasters using elec tric vehicles (EVs) by minimizing overall travel time; EVs depart from dedicated charging stations, visit all assigned evacuation shelters under battery constraints, and return for recharging. The proposed approach combines three key com ponents—a traffic congestion predicting model based on XGBoost tailored for disaster scenarios, a dynamic route update mechanism driven by real-time congestion data, and an enhanced Ant Colony Optimization (ACO) algorithm that incor porates these predictions to reformulate the vehicle routing problem. Using k-means clustering, evacuation shelters are assigned to the nearest charging stations, and predicted congestion levels are converted into expected travel times. Simu lation results with actual congestion data from the Ichihara City in Chiba Prefecture during Typhoon 15 (September 2019) demonstrated significant reductions in travel times compared to conventional distance-based routing, with improvements observed in 71 out of 120 cases in “Occurs” phase which means the 5 days right after the disaster. These findings indicate that incorporating disaster-specific congestion predictions and real-time route updates can substantially enhance emer gency power delivery in disrupted urban environments, making this method a promising tool for future disaster response planning.

Keywords: Analytical Measurement, Modeling, System Identification and Estimation
Abstract: As a proposition of this study, I am thinking to investigate how the work environment and information can improve employee well-being. It will then be used for causal analysis with a view to be addressing for MaaS and EMCs. These include invisible communication relationships. These efforts were made to visualize and activate classified questionnaires and other surveys. An approach was developed to visualize the degree of activation using structural equation modelling (e.g. SEM) based on a Bayesian approach to the data of the choices. In addition, validation was carried out to develop an approach to visualize the degree of activation from the data; visualization of means between variables based on SEM and multi-group analysis were also considered. The paper shows how important pre-processing is in SEM prediction. Also, as the previous report was able to report the main results obtained when SEM was carried out on questionnaire data extracted according to Bayesian methods, the present paper proposes whether the study is also feasible in terms of spatial construction.

Keywords: Advanced Computics, Intelligent Control, Safety, Environment and Eco-Systems
Abstract: This paper proposes a novel dispatching formulation for single-passenger micro-mobility vehicles using a quantum annealer (QA). The dispatching problem is formulated as a Quadratic Unconstrained Binary Optimization (QUBO) problem, enabling its solution through QA. Furthermore, the formulation incorporates historical usage data to achieve more efficient operation. In particular, the customer appearance frequency and destination distribution are considered by a Bayesian approach to allocate vacant vehicles to designated stations. Simulation experiments are presented to validate the effectiveness of the proposed formulation.

Keywords: Transportation Systems, Intelligent Control, Computational Intelligence
Abstract: One of the crucial tasks in realizing autonomous driving is cooperation with other traffic participants. In this study, we develop a system that generates a trajectory with speed planning to decide whether to yield, aiming to prevent deadlock and improve traffic fluidity. Although previous methods aim to cooperate with other vehicles, they have constraints requiring cloud communication with other vehicles or addressing only specific scenarios. Therefore, we propose a trajectory planning method that addresses multiple scenarios while considering interactions with other vehicles using only sensor information, without communication. Specifically, the method handles multi-modal prediction, expressing both yielding and not yielding. When the two predictions are compared, the trajectory planning is implemented towards the more agreed-upon option in the previous planning frame, ensuring a more fluid overall future traffic flow. We confirmed appropriate behavior in scenarios involving avoiding deadlock with oncoming vehicles and yielding to oncoming traffic for improved traffic fluidity through evaluations conducted with an actual vehicle.

Keywords: Robotic and Automation Systems

Keywords: Robotic and Automation Systems, Mechatronics Systems, Rescue Systems
Abstract: In a nuclear power plant, many complex pipes with an inner diameter of 20 mm or less are used and they are contaminated with high radiation. As an inspection robot for these critical condition pipes, a robot must be not only flexible and slim, but also inexpensive and easy disposable as a nuclear waste. In this study, the production method of a slim Extension type Flexible Pneumatic Actuator (EFPA) that was made of Nylon mesh sleeve and rubber tube was proposed. As a pipe holding actuator, a simple coiling motion actuator (CMA) using slim EFPA and one string was proposed and tested. A slim pipe inspection robot using the EFPA connected two CMAs from both sides was also developed. The system that can generate a delay motion for serial connected actuators by using simple orifice was also investigated. As a result, we confirmed that the tested robot was enable to pass through the bent pipe with an inner diameter of 20 mm by only giving the on/off valve operation

Keywords: Agricultural and Bio-Systems, Robotic and Automation Systems, Mechatronics Systems
Abstract: In recent years, the demand for robots to replace humans in agriculture has been increasing due to the decline in the number of farmers and an aging farming population. However, as agricultural crops grow, their size and weight change, making it difficult to automate the process; thus, automation has not been widely adopted. In this study, we focused on cabbage, a relatively heavy agricultural crop. When harvesting agricultural crops, damage to the object must be avoided. Conventional robotic hands made of hard materials can damage cabbage. To solve this problem, we developed a harvesting robot equipped with fingers based on McKibben-type artificial muscles. The fingers bend along the cabbage surface and can grasp cabbages varying in shape and weighing as much as 2 kg without damaging them. In addition, the robot can instantly grasp the cabbage because it is controlled by air pressure. The usefulness of the machine was demonstrated by developing a system in which cabbages individually cut out by humans were recognized using a depth camera and carefully harvested by the robots.

Keywords: Rescue Systems, Social Systems
Abstract: In the event of natural disasters such as earthquakes, people are often trapped under collapsed buildings, and delays in their detection and rescue can have fatal consequences. To address this issue, technologies that can accurately and quickly estimate the location of buried victims on-site are urgently needed. This research aims to miniaturize and lighten a victim-detection system that utilizes two dipole sound sources and one sound mic with a LPWA (Low Power Wide Area, 920MHz) IoT for buried human detection. We are currently planning to verify the system's positioning accuracy and practical usability in a simulated rubble environment. Based on these findings, we will refine the system to meet real-world needs better. In the future, we envision implementing the system in disaster-response platforms such as drones or mobile vehicles, ultimately contributing to labor-efficient and highly effective search and rescue operations in disaster-affected areas.

Keywords: Robotic and Automation Systems, Rescue Systems, Mechatronics Systems
Abstract: This study presents the development of BRAINS (Base containeR with open design Adaptive INtegration System), a distributed robotic system for modular combination and separation using general-purpose transport containers. To address the increasing severity of natural disasters, the system enables flexible reconfiguration of modular robots, combining work (upper body) and mobility (lower body) modules to adapt to various disaster scenarios. In this work, we demonstrated a field experiment using a scale model of a robot system intended for actual disaster deployment. In an outdoor disaster-simulated environment, the system autonomously selected between two types of mobility mechanisms based on terrain conditions and switched task modules depending on mission demands. The robot successfully adapted to tasks such as transporting exploration robots and performing drainage operations. These results validate the practicality of dynamic module selection and reconfiguration in uncertain and changing outdoor conditions. Future work will focus on quantitatively evaluating the effectiveness of dynamic reconfiguration. To clarify the advantages of our approach, we will use simulations to compare the robustness of work efficiency between our reconfigurable robots and conventional ones under dynamic environmental and task changes.

Keywords: Mechanical Systems Control
Abstract: The cybersecurity of networked control systems is crucial. As these systems have advanced, the risk of cyberattacks targeting them has grown. Developing a security-enhancement method is necessary to protect networked control systems against various cyberattacks. This study aims to propose an encrypted control system using updatable keyed-homomorphic encryption to enhance security. Updatable keyed-homomorphic encryption is an encryption scheme that adds algorithms for updating key pairs and ciphertexts to conventional keyed-homomorphic encryption. In the proposed encrypted control system, Attackers cannot obtain time-series signal data even if they succeed in cracking a ciphertext at present, because key pairs are updated at every sampling period. In addition, the proposed encrypted control system can detect not only falsification attacks but also replay attacks. To investigate the impact of encryption and the feasibility of detection against replay attacks, this study conducts stage positioning control experiments and attack detection tests using an industrial linear stage. The experimental results show that the feasibility of the encryption scheme for maintaining equal control performance to an unencrypted control system, and that the proposed control system can detect replay attacks in real time.

Keywords: Modeling, System Identification and Estimation, Identification and Estimation
Abstract: In many cases, designing suitable controllers hinges on the availability of system parameter knowledge and state measurements. Unfortunately, it is challenging to obtain either of these accurately in practical scenarios. There have been several efforts to design controllers considering the absence of parametric knowledge and/or state information. Adaptive observers are a useful tool to solve this challenging problem since they carry out simultaneous state and parameter estimation using only the input and output data. While there exist many adaptive observer designs for continuous-time systems, the design for discrete-time systems is limitedly explored. In this poster, we propose an adaptive observer for single-input single-output discrete-time linear time-invariant systems using an initially exciting (IE) input. The proposed method uses two layers of filters and a normalized gradient descent-based update law for learning the state and parameters. Unlike a persistently exciting signal which requires sufficient energy for all time, the IE signal carries limited energy for a finite time until a condition on the regressors is met. Further, this condition is online verifiable and does not depend on future values of the signal. The proposed novel IE-based adaptive observer guarantees convergence of the state and parameter estimates to their true values.

Keywords: Adaptive and Optimal Control
Abstract: In this research, we proposed a noval approach for optimizing the parameterized positive linear systems. While exist methods developed for more general linear systems can be adapted to positivity constraints,recent progress in linear programming has revealed superior conciseness and computational efficiency. However, neither the existing linear matrix inqualities method or linear programming method often cannot simultaneously grant full design freedom to the system matrices and guarantee reliable convergence to global optimal solutions. To address this issue, we propose a DC programming-based approach that allows each matrix entry to be adjusted independently, maximizing design flexibility, while obtaining an better performance on convergence rate on global solution and computation efficiency for large-scale control problems.

Keywords: Adaptive and Optimal Control
Abstract: This research focuses on clusters of atomic clocks that form a network and are controlled to synchronize their time with each other. This process achieves high-precision time synchronization throughout the network. Highly accurate time synchronization enables low-latency communications, accurate timestamping, and efficient use of communication frequencies. In addition to synchronization control, a target synchronization time is broadcast, and control is applied to track this standard time. These control gains are optimized based on the network structure. However, when the network structure changes midway through the process, the control becomes suboptimal and synchronization between the cluster clocks and tracking of the standard time is lost. In this study, the control gains are modified step by step by PSO, and control is applied in such a way that it is not optimal but better.

Keywords: Modeling, System Identification and Estimation, Robust Control
Abstract: In model-based control system design, a controller is designed using a nominal model of the plant. Therefore, the performance of the controller depends on the accuracy of the plant model, and model uncertainty can significantly degrade control performance. One approach to addressing this issue is the Model Error Compensator (MEC). MEC introduces a minor feedback loop into an existing control system to compensate for uncertainties, thereby aligning the dynamic behavior of the actual plant with that of the nominal model. This study focuses on the data-driven design of MEC and tackles the fundamental question: what type of data contains the necessary and sufficient information for designing an MEC that achieves the desired control performance? In particular, we clarify the necessary and sufficient conditions for data informativity that enable the design of an MEC filter capable of stabilizing the control system. Furthermore, we demonstrate its effectiveness through numerical examples.

Keywords: Robotic and Automation Systems, Medical and Welfare Systems, Mechatronics Systems
Abstract: We propose a magnetic control system for four-degree-of-freedom (4-DOF) actuation of an untethered permanent magnet, designed for minimally invasive applications such as intravascular navigation. The hybrid magnet array, composed of permanent magnets and electromagnets, is optimized to produce a magnetic field that generates a force on the target magnet along the gravitational (z) axis at the home position, while establishing a potential well in the directions perpendicular to gravity (x, y), as well as in the azimuthal (ϕ) and inclination (θ) rotational directions. This field configuration ensures passive stabilization in the x, y, ϕ, and θ directions. By controlling only the position along the z-axis via coil current, the relative position between the array and the target magnet can be maintained. As a result, 4-DOF control of the target magnet is achieved by simply translating the array in the x, y, z, and ϕ directions. This approach simplifies the control process to a single-axis current input while leveraging permanent magnets to reduce the required current in the electromagnets, thereby improving energy efficiency. Additionally, the open structural design offers a larger workspace than conventional systems. Simulation results demonstrate the feasibility and effectiveness of the proposed system, highlighting its strong potential for future use in magnetically actuated intravascular robotic applications

Keywords: Adaptive and Optimal Control, Intelligent Control, Robotic and Automation Systems
Abstract: This paper proposes a path planning method for automatic parking with generating a switching point by formulating a nonlinear MPC problem, where the forward and backward motions are simultaneously considered. The forward and backward motions are composed into a whole kinematic model, and the condition for the switching point that the forward and backward trajectories coincide with each other is integrated as the termination condition into the evaluation function. Especially, this paper includes the steering angle as a state variable into the kinetic model, which leads to continuously and smoothy behavior of the steering. Furthermore, this paper also study designing the weights in the evaluation function in terms of the parking behaviors of the vehicle. We will show some considerations on the weight values, which lead to generating more natural parking motions like human behaviors in the several parking situations.

Keywords: Process Control, Computer Aided Design
Abstract: This research aims to expand application targets of the data-driven prediction method based on the FRIT (Fictitious Reference Iterative Tuning) to nonlinear systems by introducing Reservoir Computing (RC). We propose a new method for efficiently designing and tuning controllers for nonlinear systems while checking simulated response.

FRIT is a method for obtaining a controller that achieves a desired response from a set of experimental data, and data-driven prediction based on FRIT estimates the closed-loop characteristics when a specific controller is used. This approach provides an a priori idea of the performance limits achievable within the structural constraints of the controller. However, in the conventional approach, the closed-loop characteristic model to be estimated is linear, making it difficult to predict the response of a controlled system with nonlinearity.

In this study, we employ RC, which has excellent nonlinear time series data representation capability, to estimate this closed-loop characteristic model. RC requires small learning computation cost despite being a nonlinear model, because it fixes the weights of internal states (reservoirs) and only adjusts the weights of output layers. This characteristic is particularly useful in controller tuning situations where it is necessary to repeatedly predict the response of a large number of controller candidates.

In this paper, we verify the effectiveness of the proposed method through numerical simulations.

Keywords: Mechatronics Systems, Medical and Welfare Systems, Mechanical Systems Control
Abstract: With the progression of an aging society, the number of elderly individuals and disabled people with severe physical disabilities is steadily increasing, leading to a growing burden on caregivers. In light of the need to enhance Quality of Life (QOL), welfare devices in the field of rehabilitation are becoming increasingly important. In particular, the development of soft actuators that are gentle on the human body is in high demand. In this study, we propose a Spring-Supported Flexible Pneumatic Actuators (SSFA) that offers vertical compliance and various actuator motions (stretching, rotating and bending). SSFA consists of 12 Extension type Flexible Pneumatic Actuators (EFPAs) and a coil spring. First, the static characteristics of the actuator, which is the relationship between EFPA pressure and actuator posture (bending angle, central displacement, etc.), were investigated through both experimental testing and mechanical analysis and an analytical model was constructed. Furthermore, from the observation of the actuator displacement when applying an impact load to the SSFA and removing the load, we demonstrated that the SSFA structure exhibits adaptability to external forces. These results suggest that the SSFA has a potential for application in rehabilitation devices, particularly in scenarios where flexibility, adaptability, and safety are required.

Keywords: Mechatronics Systems, Medical and Welfare Systems, Mechanical Systems Control
Abstract: To develop a small , lightweight, and inexpensive control valve is an innovative technology that can support fluid power system to drive soft mechanisms. In the previous study, a low-cost servo valve using a slide gate mechanism and diaphragm was successfully developed. In this study, as an application of multi-port control valve using slide gate mechanism, the compact pneumatic control system using the valve for a flexible pneumatic spherical actuator (FPSA) that consists of six extension type flexible pneumatic actuators (EFPAs) was proposed and tested. The control system consists of FPSA, the tested switching valve and two On/Off solenoid valves for supply and exhaust. The attitude measuring system of FPSA by measuring the center length of FPSA using a wire type linear potentiometer was also proposed and tested. As a result, we confirmed that FPSA could perform successfully the rotational motion with constant bending angle, even if the compact control system with less control inputs was used.

Keywords: Robotic and Automation Systems, Modeling, System Identification and Estimation, Biological and Physiological Engineering
Abstract: This study introduces a novel cable-driven exoskeleton designed for gait rehabilitation studies in rats. The exoskeleton assists all three hindlimb joints of the rat (hip, knee, and ankle) while ensuring proper interjoint coordination and the natural quadrupedal posture. This assistance is realized through a 2-DOF bar mechanism that emulates the biomechanics of rats. Engineered to be compact, lightweight, backdrivable, and sufficiently powerful, the proposed system minimizes physical stress on the animal while allowing a wide range of assistive forces to be applied. These features are achieved through a combination of a cable power transmission system and direct-drive motors positioned outside the exoskeletal structure. The desktop experiments demonstrated that the exoskeleton could precisely replicate the rat’s kinematic gait patterns and remain backdrivable whether powered or unpowered. The feasibility of gait assistance was further confirmed in an anesthetized rat, where synergistic gait patterns were observed between the joints. Hence, the system holds the potential to enable controlled comparative neurorehabilitation studies in rats. These studies can help unveil neural recovery mechanisms and design optimal exoskeleton control strategies for rehabilitation in humans.

Keywords: Intelligent Systems, Human-Machine Systems, Human Interfaces
Abstract: While meetings are essential for decision-making and information sharing, lengthy meetings, derailing discussions, ending without concrete conclusions, concentrating opportunities to speak on specific participants, or the presence of non-participants lead to the waste of valuable time and resources, and also reduce employee motivation. Improving the quality of discussions in meetings is important not only to save time, but also to improve the overall performance of the organization. To realize more meaningful discussions, discussion analysis and support based on objective data are required. In this study, actual meeting data is collected and discussions are quantitatively analyzed using methods such as morphological analysis, correspondence analysis, and topic analysis. The analysis visualizes the level of activity of discussions and the transition of discussions by time series. By feeding back the results of these analyses to the participants, the tendency to speak and the overall atmosphere of the meeting are objectively recognized, promoting the activation of discussions. In addition, based on the analysis results, a large-scale language model (LLM) is used to provide participants with specific advice and measures to improve discussions, to support more effective discussions.

Keywords: Robotic and Automation Systems, Mechatronics Systems, Autonomous Decentralized Systems
Abstract: This talk proposes an approach to enhance adaptive and learning capabilities in real-world environments by integrating Physically Embodied AI into bio-inspired robotics. Specifically, the research experimentally investigates odor source localization using a hexapod robot inspired by insect olfactory navigation. We demonstrated that by combining adjustments to the robot's body property (body height and odor intake volume) with a simple search algorithm, it is possible to detect and track odors more efficiently, demonstrating the effectiveness of ``embodied intelligence'' achieved through cooperation between the body and AI.

Keywords: Biological and Physiological Engineering, Computational Intelligence
Abstract: Insects exhibit highly adaptable rhythmic locomotion by flexibly modulating both inter-leg and intra-leg coordination in response to environmental and contextual perturbations. While previous studies have investigated the neural mechanisms underlying rhythm generation using phase response curves (PRCs), most have been limited to simplified paradigms—such as single-leg stepping or coarse mechanical disturbances—that do not capture the full complexity of natural locomotion. This study introduces a novel experimental framework called ``Motion Hacking", which enables temporally precise interference with motor control in freely walking stick insects ({it Carausius morosus}) through targeted electrical stimulation of individual leg muscles. This methodology allows for the quantification of phase-dependent responses to perturbations under semi-natural walking conditions and facilitates the examination of cascading coordination effects across legs. Using PRC analysis, we explored how stimulation of protractor and retractor muscles influenced the locomotor rhythm as the insects walked on slippery surface ground.

Keywords: Biological and Physiological Engineering, Integrative Biophysical Engineering and Informatics
Abstract: Soft actuators rely on external fluid pressure supplies for operation. For soft robots to function autonomously, they require an integrated fluid pressure supply system; however, no existing system is suitable for deployment in unstable environments. In this study, we drew inspiration from the pump systems of living organisms, which have evolved to function across diverse environments. The heart, as a natural fluid pump, plays a crucial role in circulation, but quantifying the relationship between its morphology and pumping function is challenging due to the interaction of structural and biochemical factors. Here, we focused on structural variations in frog hearts, fabricating heart models with different membrane thickness to assess their functional implications. Our findings revealed that membrane thickness directly influences diastolic function, and this result was consistent with biological data. From this, we concluded that the method we developed can be used as a system to evaluate the morphology and function of the heart, providing a foundation for bioinspired fluid pressure systems in soft robotics.

Keywords: Intelligent Systems, Autonomous Decentralized Systems, Robotic and Automation Systems
Abstract: Sensory modalities used for animal orientation, such as vision, audition, and olfaction, differ in their physical properties, which may have influenced the evolution of distinct behavioral strategies for each sense. However, it remains unclear whether animals use separate strategies for each sensory modality or share a common goal-oriented strategy. This study focuses on odor-, light-, and sound-source orientation behaviors in insects and develops a closed-loop system that integrates an odor- or light-sensing mobile robot with the auditory navigation (phonotaxis) of a cricket. The system consists of a mobile robot that acquires sensory signals from the real world and a cricket placed in an auditory virtual reality, which remotely controls the robot in response to substituted auditory stimuli. By substituting the auditory input of a cricket with visual or olfactory cues, this system provides a direct method for investigating whether a common strategy underlies goal-directed behaviors across different sensory modalities.

Keywords: Biological and Physiological Engineering, Integrative Biophysical Engineering and Informatics, Autonomous Decentralized Systems
Abstract: In this study, focusing on the silkmoth’s odor source behavior, we developed a system for simultaneous odor dynamics and behavior measurement to investigate the effect of silkmoth wing flapping and wing structure on odor reception. We used a chemical mixture of smoke and pheromone (Bombykol) to visualize the odor. In the simultaneous measurement experiment of odor dynamics and behavior, two wind conditions and four silkmoth wing conditions were used, where the wings were excised to modify the physical characteristics. Our findings revealed that the silkmoth with only the forewing can localize the odor source with high probability, suggesting that the forewing has the effect of odor intake. Additionally, our results suggested that wing flapping creates intermittency and direction in odor reception.

Keywords: Robotic and Automation Systems, Autonomous Decentralized Systems, Rescue Systems
Abstract: Insects have excellent odor-searching capabilities using their movable antennae and arrays of odor-receptive cells to locate odor sources. In previous studies on odor-searching robots with insect-antenna-inspired gas sensor arrays, the antenna movement was limited to horizontal movement, restricting odor detection to that plane. In this study, we focused on vertical and horizontal movements to allow the robot to detect odors at different heights. The artificial antenna, equipped with three gas sensors on each side, was actuated in both the horizontal and vertical directions using two servo motors. We implemented a parallel linkage to the antenna to improve the response consistency of the sensors that exhibited strong directional sensitivities. We demonstrated that the three-dimensional movement of the antennae enabled the detection of odor sources located above the robot. Furthermore, we demonstrated that the three-dimensional orientation of the artificial antenna captured the odor distribution in three-dimensional space. These results suggest that the three-dimensional movement of the antennae enhances the odor-searching capability of mobile robots.

Keywords: Autonomous Decentralized Systems
Abstract: Gallop is the fastest gait employed by quadruped animals, with variations in flight timing which characterize periods when all limbs are off the ground.While previous studies have investigated gallop dynamics with various flight timings using analytical or optimization approaches, the underlying control mechanisms that determine these flight timings remain insufficiently explored.In this study, we investigate the control mechanisms underlying galloping gaits through mathematical modeling and simulations.Our results showed that the strength of the head and trunk actuation serves as a determinant for flight timing variations in gallop gait.

Keywords: Autonomous Decentralized Systems, System Engineering, Intelligent Systems
Abstract: Centipedes, despite lacking advanced vision, can adaptively navigate complex environments through active and exploratory sensing using their antennae and the anterior part of their body. This study aims to construct a mathematical model that integrates environmental sensing and autonomous decentralized locomotor control to understand the inherent mechanisms of this exploratory navigation.

Keywords: Adaptive and Optimal Control, Modeling, System Identification and Estimation, Intelligent Control
Abstract: This study introduces a Linear Matrix Inequality (LMI) approach for adaptive Linear Quadratic Regulator (LQR) based on Q-learning to address discrete-time systems under both deterministic and stochastic cases. The proposed method focuses on estimating unknown matrices that generate the Q-function using LMI-based constraints. Main advantage of the LMI-based approach lies in its ability to guarantee the positive definiteness of estimated matrices. In stochastic cases, the LMI formulation realizes the bias-free estimation without instrumental variables, which is indispensable for the bias-free estimation by way of the regressor vector consisting of quadratic basis functions. The proposed method introduces how the semi-definite programming (SDP) can be solved for the estimation of the Q-function. Simulation results show that the proposed method improves estimation accuracy and enhance control performance, demonstrating the effectiveness of the proposed LMI-based design.

Keywords: Adaptive and Optimal Control, Robust Control, Modeling, System Identification and Estimation
Abstract: This study proposes a design method for robust regulator optimization using a fractional-order controller. Fractional-order systems may be more suitable for control systems than integer-order systems due to their greater design flexibility. However, compared to integer-order control systems, the design parameters of fractional-order control systems are more numerous and complex. In this study, fractional-order control systems are designed directly from the controlled data. Therefore, by utilizing the proposed method, fractional-order control systems can be designed relatively easily.

Keywords: Adaptive and Optimal Control
Abstract: In recent years, the cooperative transportation of objects by multiple robots has attracted attention. In this study, we propose a method for object transportation by multiple agents that combines formation control and optimization of reference force in consideration of grasping force constraints. By optimizing the reference force at each time step, grasping forces can be adaptively generated even when the object rotates. Through simulations, we demonstrate that the object can be successfully grasped and transported by optimizing the reference force.

Keywords: Adaptive and Optimal Control
Abstract: This paper proposes an improved vehicle path-following control method using an adaptive output feedback system with a radial basis function (RBF) neural network for feedforward input. The approach enhances a previous method based on the Almost Strictly Positive Real (ASPR) property and parallel feedforward compensation (PFC), which suffered from tracking errors. The new method introduces RBF to the feedforward input to improve accuracy in skidding angle tracking. A simplified two-wheel vehicle model is used to design and analyze the controller. Stability is proven through Lyapunov theory, showing boundedness of system signals and error convergence. Numerical simulations demonstrate that the proposed method reduces tracking errors compared to the conventional approach, especially in transient situations during directional transitions. The results confirm the effectiveness of integrating RBF into adaptive control to handle model uncertainties and varying road conditions, contributing to safer and more comfortable autonomous vehicle operation.

Keywords: Adaptive and Optimal Control, Intelligent Control, Modeling, System Identification and Estimation
Abstract: PID parameter tuning has traditionally relied on the expertise of the tuner, requiring an understanding of the effects of proportional, integral, and derivative actions. However, in practice, users often possess control performance requirements without having such specialized knowledge. On the other hand, generative AI, exemplified by ChatGPT, has made remarkable progress and can emulate human trial-and-error behavior depending on prompt design. It can also acquire PID tuning know-how from the vast amount of information available on the web. In this study, we propose a method for obtaining PID parameters that meet control performance requirements using generative AI, even in the absence of expert knowledge. Furthermore, we discuss the potential application of this method to interactive educational simulators, where learners can experience control tuning through dialogue with AI. This research provides a foundation for bridging control engineering education and AI-driven instructional tools.

Keywords: Adaptive and Optimal Control
Abstract: This paper deals with the control problem of an engine intake and exhaust system with low-pressure exhaust gas recirculation (LPEGR). Generally, engine intake and exhaust controls have been performed using feedforward control or PID control, based on experimental results. However, in recent years, with the introduction of LPEGR to reduce exhaust gas emissions, the control system has become more complex, making control more difficult than ever before. In this paper, we propose an output predictive control system for the engine air-path system with EGR to enhance control performance in the presence of EGR transfer delay. By using the predicted input for a delay time ahead, the proposed method can improve the control performance against EGR transfer delay. The performance of the proposed control method will be confirmed through numerical simulations for an engine intake and exhaust system model with EGR time delay.

Keywords: Modeling, System Identification and Estimation, Nonlinear Control, Adaptive and Optimal Control
Abstract: Networked control systems are exposed to a variety of threats. In repeated replay attacks, malicious users steal and record signals from control systems and play them back repeatedly. Hence, such attacks are difficult to detect and can be fatal to the control systems. This paper proposes a new detection of repeated attacks using an integrator with a piecewise constant auxiliary signal. Theoretical analysis shows reliable detection using the proposed method. Moreover, numerical simulation evaluates its effectiveness.

Keywords: Adaptive and Optimal Control, Guidance and Flight Control, Mechanical Systems Control
Abstract: This paper proposes a novel event-triggered adaptive control of a drone-ball-beam (DBB) system, in which a drone and a ball-beam system are connected by wires. The proposed controller can track the ball position to a reference signal even under parameter variations. Furthermore, by introducing an integrator with an auxiliary signal into the control system, communication failures between the controller and the drone can be detected. The effectiveness of the proposed control system is evaluated by theoretical analysis and numerical simulation.

Keywords: Modeling, System Identification and Estimation, System Engineering
Abstract: In this paper, we present a novel prediction technique based on Kriging methods. Here, instead of assuming that the dynamics of the system are completely unknown and build a complete black-box model of the system, we opt to inject previous knowledge of the system into the black-box model in order to improve the performance of the obtained predictor. This can be easily done by adding new constraints to the Kriging predictor. These can be added in a hard-constraint or a soft-constraint manner, obtaining different results depending on this choice. Also, embedding physics-based conditions into the predictor is compatible with both the local-data based Kriging and the kernel-based Kriging technique, allowing for a flexible modeling depending on the application. Finally, the proposed predictors were tested against their state-of-the-art versions to show their effectiveness in a numerical example.

Keywords: Modeling, System Identification and Estimation, Multivariable Control, Intelligent Control
Abstract: This study investigates graph reconstruction methods for multi-agent systems based on structural controllability. In particular, this paper proposes a procedure for recovering the structural controllability in the presence of failures in the communication links and some follower agents. Numerical simulations of formation control confirm the effectiveness of the recovery procedure.

Keywords: Intelligent Control, System Engineering
Abstract: The paper demonstrates numerically and experimentally for mobile robots that select the minimum feedback vertex set (FVS) as leaders in leader-follower consensus control of single integrator dynamics to minimize the convergence rate. Formation control using a virtual structure for unicycle kinematics on the plane illustrates that leaders based on the minimum FVS exhibit fast convergence. In the experiment, formation control of parallel two-wheeled mobile robots also illustrates the effectiveness of selecting the leaders as the minimum FVS.

Keywords: System Engineering, Robotic and Automation Systems, Intelligent Control
Abstract: The swarm robotics system (SRS) uses multiple robots inspired by the self-organizing abilities of social organisms such as ants and bees. One of the methods for constructing SRS control systems is the evolutionary robotics (ER) approach, which automatically designs controller parameters. The ER approach is a method for automatically creating the artificial neural network controller parameters that often deal with homogeneous SRS. This study aims to investigate heterogeneity via perceptual variations between robots, primarily enabled by input acquisition sensors, which play an essential role in the acquisition of swarm behavior. Heterogeneous robots with different perceptual capabilities outperform homogeneous ones.

Keywords: Adaptive and Optimal Control, Robotic and Automation Systems, Multivariable Control
Abstract: Persistent coverage control is known as one of the control methods which enables efficient and persistent monitoring by a group of mobile robots. In typical persistent coverage control, a time-varying density function is utilized to represent the state of monitoring within the region, which is then used to create the control inputs. However, typical persistent coverage control methods assume that the environment is known, and thus they do not work well in environments with unknown obstacles. In this paper, we derive a design method of persistent coverage control in unknown environments which simultaneously estimates unknown obstacles by Gaussian process regression. The performance of the proposed method is discussed based on simulation and experimental results.

Keywords: Nonlinear Control, Mechanical Systems Control, Autonomous Decentralized Systems
Abstract: For circular path following of two-wheeled mobile robots (WMRs) with a specified circulating velocity in a desired formation pattern, this paper proposes a decentralized state feedback controller by combining line-of-sight (LOS) guidance and consensus control mechanisms. The theoretical guarantees along with assumptions are presented. First, stability of the LOS guidance is established uniformly over forward velocities of WMRs by constructing a Lyapunov function. Then, it is combined with the consensus control determining the forward velocities using stability theory of stable nonlinear systems in cascade. The validity and usefulness of the designed controller are demonstrated through simulation and experimental results.

Keywords: Mechatronics Systems, Robotic and Automation Systems, Mechanical Systems Control
Abstract: Multi-robot systems are a promising technology for their flexibility, adaptability, and efficiency. These systems enable autonomous coordination and can complete complex tasks through collective behavior. This research introduces a modular multi-robot system for shape formation, highlighting mechanical design and command sequencing to achieve coordinated movements. Each robot, built as a square-shaped module, features a motor-driven rack and pinion mechanism along with electromagnets. These features allow individual robots to push, connect, and stabilize with other robots in a three-dimensional coordinate system. The system uses sequential commands to ensure synchronized actions among the robots and alignment during movement. Experiments were conducted in a controlled simulation environment to test the reliability of the rack and pinion mechanism, the stability of the electromagnet connections, and the precision of coordinated movements. Experimental results demonstrate the ability of robots to construct simple 3D shapes. These experiments highlight the system's effectiveness in constructing various structures, with potential applications in automated construction, modular robotics, and interactive entertainment. The findings suggest that such systems can enhance scalability and efficiency in tasks requiring shape formation, contributing to future advancements in robotics and automation by opening new research and application methods in multi-robot systems.

Keywords: Nonlinear Control, Robotic and Automation Systems, Autonomous Decentralized Systems
Abstract: This paper proposes a distributed control barrier functions (CBFs) approach to guarantee field-of-view overlap for cooperative visual localization in multi-agent systems. We formulate the problem on the Special Euclidean Group SE(3) and introduce a probabilistic CBFs that evaluates the visibility of feature points to ensure that multiple agents maintain shared visual features. The proposed method incorporates a distributed optimization algorithm based on the primal-dual method of multipliers (PDMM), enabling each agent to compute control inputs locally while satisfying global safety constraints. Simulation results demonstrate the effectiveness of our approach in both centralized and distributed implementations, showing that agents can maintain shared field-of-view while following desired trajectories.

Keywords: Robotic and Automation Systems, Intelligent Control, Autonomous Decentralized Systems
Abstract: Modular robotic systems are essential in addressing complex challenges across various applications, including manufacturing, logistics and exploration. Their ability to adapt and reconfigure makes them a promising solution for rapidly evolving operational environments. The self-reconfigurable nature of these systems allows individual robotic units to autonomously form specialized structures, which opens innovative applications, particularly in shape formation tasks. This research aims to develop a structured algorithm for 3D shape formation within modular robotic systems. The primary goal is to improve operational efficiency and adaptability by optimizing the movement coordination of robotic units as they transition into desired configurations. The developed algorithm integrates control strategies with mechanical design principles, allowing each modular unit to utilize sliders for movement and electromagnets for stabilization. The algorithm systematically evaluates the current configuration against a target arrangement, generating coordinated movement sequences. The algorithm's validation through simulations demonstrated that it enhances the efficiency of shape formation tasks. Multi-step planning reduces redundant movements and enhances coordination, resulting in practical advantages such as improved stability and accuracy during reconfiguration.The findings highlight the effectiveness in addressing the complexities associated with modular robotic systems.

Keywords: Nonlinear Control, Robust Control, Multivariable Control
Abstract: This paper proposes the design of time-synchronized control for nonlinear affine systemsl with disturbance. Since the existing design uses super-twisting disturbance observer, it requires the known upper bound for the time- derivative of disturbance as well as additional conditions for the observer gains. This leads to the complexity and difficulty in the design, which suggests the need of more practical design. Thus, practical time-synchronized stability is defined to ease the design and a practical time-synchronized control system is given to achieve the objective. We compare the proposed design and existing design by a numerical simulation.

Keywords: Mechanical Measurement, Sensors and Transducers, Signal and/or Image Processing
Abstract: Conductive sponge sensors have attracted growing interest for their high sensitivity, flexibility, low cost, and simple fabrication, making them ideal for wearable electronics and posture monitoring. However, optimizing performance requires selecting suitable materials and electrode configurations. This study compares conductive sponge sensors using three electrode materials—aluminum, copper, and carbon rubber—for back posture detection. Sensors were evaluated based on mechanical properties (compressive stress and strain) and capacitance responses. Capacitance values ranged from 0–9.5 F for aluminum, 0–10.0 F for copper, and 0–7.5 F for carbon rubber. Environmental tests under cyclic loading (up to 30 cycles) confirmed excellent stability and minimal signal fluctuation. Additionally, fast response times were observed under dynamic conditions, such as loading and unloading a 5 kg weight, demonstrating reliability in real-time applications. The findings suggest copper electrodes offer the highest capacitance, while all configurations provide reliable performance. This study offers useful insights into material selection for conductive sponge sensors and supports their integration into posture monitoring, healthcare diagnostics, environmental sensing, and future wearable systems.

Keywords: Signal and/or Image Processing, Transportation Systems, Computational Intelligence
Abstract: This paper presents a simplified training strategy for drivable area detection, focusing on road environments where lane markings are missing or unclear. Based on the YOLOP architecture, the proposed method removes the lane detection branch and adopts a two-stage training process. The model is first trained on object detection to allow the encoder to learn useful visual features. In the second stage, segmentation is introduced, and both tasks are trained together in a multitask setup. This helps the model maintain detection performance while learning to identify drivable areas accurately. This design reduces task interference and helps the model better adapt to unstructured roads. The approach aims to improve model accuracy and robustness while maintaining architectural efficiency for real-world applications. Evaluation on a custom dataset featuring unstructured road conditions demonstrated that the proposed model achieved notable segmentation accuracy (92.6% mIoU) and reliable object detection (91.3% recall, 68.1% mAP@0.5).

Keywords: Robotic and Automation Systems, Multivariable Control
Abstract: This paper proposes a design of model predictive control (MPC) for an omnidirectional mobile robot. The robot is modeled as a linear time-invariant (LTI) system and constrained to holonomic motion without rotation and operates in dynamic environments that contain moving and static obstacles. A virtual potential based on a linearized vector is introduced to emulate the influence of obstacles, allowing obstacle avoidance without modifying the robot model. The simulation results show that the proposed approach achieves a minimum obstacle clearance, an average tracking error of 0.3099 m, and a 60% reduction in computation time compared to nonlinear MPC.

Keywords: Temperature Measurement, System Engineering
Abstract: The demand for air conditioners continues to rise, especially in high-temperature environments where their efficiency depends heavily on ambient conditions. As a result, air conditioners have become a significant source of energy consumption, leading to higher electricity bills. To address this issue, ground source heat pumps (GSHPs) provide cooling by transferring heat into the ground through ground heat exchangers (GHEs), relying on stable ground temperatures for optimal efficiency. This study is the first in Thailand to implement a hybrid ground source heat pump (HGSHP) in cooling mode by integrating an inverter-type air conditioner with a GSHP that utilizes the energy piles of a green academic building as a heat sink. In this cooperative HGSHP system, only a portion of the heat extracted from indoor spaces is transferred to the ground, while the remaining portion is discharged into the ambient air. In contrast, traditional systems, known as air-source heat pumps (ASHPs), discharge the entire heat load into the ambient air. Our HGSHP system enhances overall heat transfer efficiency compared to traditional ASHPs. The results indicate an energy reduction of 13.10% in compressor workload. Additionally, the integration of energy piles during building construction lowers installation costs and operational time associated with borehole drilling. This study underscores the potential of utilizing GSHPs with energy piles in short-term experiments to improve air conditioning system

Keywords: Robust Control, Mechanical Systems Control
Abstract: This paper proposes a hybrid control strategy that combines sliding mode control (SMC) and field-oriented control (FOC) for speed tracking of a brushless direct current (BLDC) motor. The outer loop, based on using SMC, enables fast response to load torque and external disturbances, while the inner loop, based on FOC with proportional–integral (PI) controllers, ensures precise control of torque and current. In the proposed hybrid SMC-FOC strategy, a fractional integral terminal sliding mode control (FITSMC) is selected, since it achieves faster and more accurate convergence while avoiding the need for unbounded control effort near the origin. Moreover, space vector pulse width modulation (SVPWM) is utilized to enhance system efficiency further and reduce torque ripple. This hybrid SMC-FOC strategy was tested on a simple, low-cost ESP32 microcontroller. Experimental results demonstrate improved speed response and robustness compared to the conventional PI-FOC under both symmetric and asymmetric load conditions. These results suggested that the solution is practical for applications requiring affordable yet reliable motor control.

Keywords: Robust Control, Modeling, System Identification and Estimation
Abstract: This paper presents a fractional integral terminal sliding mode control (FITSMC) for speed tracking of a DC motor employed in phenotyping robots for green environment applications. Phenotyping robots require high-precision motion control to ensure reliable and repeatable measurements of plants under field conditions where mechanical and load disturbances exist. To address these challenges, the proposed FITSMC ensured finite-time convergence of speed-tracking errors, even in the presence of unknown disturbances and modeling uncertainties. Unlike conventional sliding mode control, the FITSMC enables faster and more accurate convergence without requiring infinite control effort near the origin. Simulation results confirm the effectiveness and robustness of the proposed method, highlighting its potential for implementation in green robotic systems for precision agricultural applications.

Keywords: Robotic and Automation Systems, Intelligent Control, Human-Machine Systems
Abstract: This study proposes safe speed control for a personal mobility robot in environments with occlusion areas. We first define the potential risk of occlusion areas as the Probability of Potential Risk (PPR), which represents the likelihood of remaining in a safe state over a specified time horizon. In the proposed method, the PPR is analytically estimated using a Bayesian network and is incorporated into the speed control system. We evaluate the proposed method through simulations and real-world experiments in an indoor environment containing occlusion areas. Our experimental results shows that the proposed method reduces speed for the occlusion area and maintains a safe distance from pedestrians. Compared to simulations, delays in deceleration onset and reduced deceleration magnitudes are observed, attributed to system response delays and sensitivity settings of the safety constraint.

Keywords: Human-Machine Systems, Robotic and Automation Systems, Intelligent Systems
Abstract: It is difficult for users to drive a personal mobility while avoiding obstacles, including pedestrians, and preventing deviations from the sidewalk. Therefore, this paper focuses on driving assistance provided by the personal mobility itself. We propose a shared controller that integrates inputs from both the user and the mobility. To enable safe driving assistance, the mobility computes control commands based on the surrounding environment. To achieve this, a semantic map composed of 3D point clouds is built using online SLAM through sensor fusion of a 3D LiDAR and an RGB camera. In this map, point clouds are classified into drivable and non-drivable areas. However, for dynamic obstacles such as pedestrians, point clouds classified as non-drivable along their walking trajectories may also be included in the map. To address this issue, pedestrians are detected through point cloud clustering, and their point clouds are excluded from the online SLAM process. Driving assistance experiments demonstrate that the personal mobility equipped with the proposed shared controller can successfully navigate toward its destination.

Keywords: Human-Machine Systems, Robotic and Automation Systems, Modeling, System Identification and Estimation
Abstract: This paper proposes a novel local planning approach for Autonomous Mobile Robots (AMRs) to reduce pedestrian uncertainty during human-robot interactions in shared spaces. We introduce a quantitative measure of pedestrian uncertainty through a "decision entropy" framework that captures the uncertainty in crossing-order decisions. While existing research has focused primarily on predicting pedestrian physical movements, our method addresses the cognitive aspects of decision-making processes during robot-human encounters. By simultaneously optimizing both linear and angular velocities, our controller minimizes pedestrian uncertainty and creates more predictable robot behaviors. We extend previous decision entropy concepts to incorporate rotational movements, enabling the robot to clearly communicate its intentions through deliberate motion patterns. Experimental validation in real-world crossing scenarios demonstrates that our approach significantly reduces uncertainty metrics when the AMR follows pedestrians, resulting in more decisive interactions and improved pedestrian comfort. This work establishes a foundation for developing socially-aware navigation systems that consider pedestrian psychological responses in shared environments.

Keywords: Human-Machine Systems, Intelligent Systems, Innovative Systems Approach for Realizing Smarter World
Abstract: Human-Robot Collaboration (HRC) is a key enabler for achieving Industry 5.0, where humans and robots work together on tasks with greater safety, productivity, and efficiency. However, human intervention can slow down the robot's work speed and reduce its productivity. This study conducts a preliminary investigation into the effects of human intervention in HRC workspace, focusing on a conveyor-driven production line with varying operational speeds. The experiment workspace consisted of human and robot sorted together strawberries into three different size. The work scenario was divided into three cases-human alone, robot alone, and HRC-each tested under two conveyor speeds: standard and high. The experimental results revealed that HRC achieved the highest accuracy among the three conditions, with 91.7% and 91.3% for the standard and high-speed conveyor settings, respectively. High identification accuracy was observed at higher speeds with reduced human workload through collaboration, allowing greater concentration on the task and minimizing false identifications.

Keywords: Human-Machine Systems, Intelligent Systems, Human Interfaces
Abstract: Hugging is a fundamental form of physical communication that conveys emotions such as comfort and support. As digital communication advances, enabling remote tactile interaction has become essential for maintaining emotional closeness between physically separated individuals. This study proposes a wearable system that simulates hugging through camera-based gesture recognition and air-pressure haptic feedback. The system uses OpenPose to detect the user’s hugging motion and classifies it into light or firm hugs based on wrist position and joint angles. A jacket with embedded air pockets inflates specific regions to recreate tactile sensations. Experiments showed that the system correctly recognized intentional hug gestures with 82% accuracy and reproduced approximately 60–70% of the pressure experienced in real hugs. Misclassification patterns are analyzed, and future improvements such as pressure sensors and machine learning-based joint analysis are discussed. Currently, the system provides tactile feedback only to the hug initiator; feedback for the receiver is not yet implemented. Future work will aim to enable bidirectional interaction for a more immersive hugging experience.

Keywords: Mechanical Systems Control, Virtual Reality Systems, Entertainment Systems
Abstract: This paper presents HapPunch, a novel wrist-worn haptic device designed to passively deliver punching sensations for augmented reality (AR) and virtual reality (VR) applications. The device provides tactile feedback through a simple, passive mechanism, where a movable mass drives to a knuckle flap via mechanical linkages. The motion of the flap is activated by the inertial force generated from the user's punch while providing realistic tactile feedback in virtual interactions. Experiments were conducted using four basic boxing punches—jab, cross, hook, and uppercut—to evaluate the system's performance. Contact sensors were used to measure the time interval between the completion of each punch and the corresponding tactile feedback. Results demonstrated that the device consistently delivered tactile feedback within a time range of 0.1 to 0.3 seconds. Moreover, limitations and future works are discussing more in the research paper. the results highlight the potential of HapPunch as a practical and immersive haptic feedback solution for future VR applications.

Keywords: Robotic and Automation Systems, Intelligent Control, Human-Machine Systems
Abstract: Service robots are increasingly deployed in restaurants and hospitality venues to address labor shortages and meet demand for contactless services. However, conventional delivery robots using fully autonomous navigation often struggle in crowded spaces. They may block human paths or stop abruptly, risking payload displacement. To address these issues, we propose a cooperative control system with supervisory coordination for delivery robots in dynamic environments. In this model, humans manage path planning and obstacle avoidance, while the robot follows behind to assist with stability and safety. The system comprises three modules: a QR code-based tracking unit, a virtual-area obstacle detector using LiDAR, and a supervisor node that resolves conflicting commands from ROS2 nodes. We validated the system using a real-world robot platform (Kachaka). The proposed supervisory control was compared with manual operation via a smartphone application. Results showed a 34.1% reduction in acceleration variation, lowering the risk of payload tipping. These findings indicate that the proposed system prevents control conflicts and improves motion smoothness. It offers a practical and robust solution for service robots in human-shared environments.

Keywords: Power Systems Control, Innovative Systems Approach for Realizing Smarter World, Nonlinear Control
Abstract: In this paper, we propose a design method for inverter-based devices that satisfy passivity with respect to a lossy power system, under the assumption that the ratio of resistance to reactance is constant. The power system is modeled as a negative feedback interconnection of a transmission network subsystem and a bus subsystem. By introducing a different input-output port structure from that used for lossless power systems, we define an energy function applicable to lossy systems and analytically demonstrate passivity. Furthermore, a numerical example demonstrates that the set of passive equilibrium points coincides with the set of small-signal stable equilibrium points.

Keywords: System Engineering
Abstract: This paper addresses the day-ahead scheduling of thermal power plants and multiple storage batteries under interval predictions of photovoltaic (PV) and demand. The goal is to determine the exact feasible range of each optimal operation schedule for all possible PV/demand profiles. By leveraging the monotonicity of solutions with respect to PV/demand levels, the required evaluations can be kept finite. While previous studies focused on a single or two storage batteries, this work extends the monotonicity analysis to n batteries. We present sufficient conditions ensuring the thermal plant's optimal generating power remains monotonic, offering a broader framework for future power systems.

Keywords: Power Systems Control, Nonlinear Control, Network System Integration
Abstract: In this study, we propose a theoretical framework for assessing the small-signal stability of power systems by extending the concept of passivity from an energy function perspective. Specifically, we formulate new necessary and sufficient conditions for the small-signal stability of power systems in an analytically tractable form. Furthermore, based on this insight, we establish a systematic methodology for reducing a sub-area within the power system to an equivalent single synchronous generator model while rigorously preserving its small-signa stability characteristics. This proposal can contribute to the reduction of computational burden, elimination of conservatism, and improvement of analytical accuracy in the stability analysis of large-scale, complex power systems, thereby supporting more reliable decision-making in the planning and operation of future power systems.

Keywords: Modeling, System Identification and Estimation, System Engineering
Abstract: This research evaluates the life-cycle exergy and economic performance of a renewable hybrid microgrid system for rural electrification in Khlong Ruea, Thailand. The methodology integrates life cycle exergy analysis with economic criteria, using Net Present Cost (NPC) and Cost of Energy (COE), to evaluate system performance. The results demonstrate significant improvements in key performance metrics compared to the existing system. The Exergy Payback Time is reduced from 0.75 years in the present system to 0.64 years in the proposed system, indicating a faster return on exergy investment. Non-Renewable Exergy Use is substantially lower in the proposed system (2.56%) compared to the present system (58.2%), highlighting a greater reliance on renewable sources. The economic advantages of the proposed system are also evident. The Net Present Cost (NPC) decreases drastically from USD 3,400,000 in the present system to USD 362,687 in the proposed system. Similarly, the Cost of Energy (COE) drops from USD 0.6-0.8/kWh to USD 0.19/kWh. These improvements suggest a significantly more economically viable solution for rural electrification. A comparative analysis with other hybrid microgrid configurations reported in the literature further underscores the potential of the proposed system as a sustainable and economically attractive solution for rural electrification.

Keywords: Computational Intelligence, Intelligent Systems
Abstract: Many real-world applications involve highly expensive multiobjective optimization problems (MOPs), where the number of allowable function evaluations is severely limited to only a few dozen. This paper proposes a surrogate-assisted multiobjective evolutionary algorithm (SAMOEA) specifically designed for such challenging scenarios. We introduce a novel infill criterion that selects most effective candidate solutions by rigorously evaluating their contributions to both convergence and diversity, and integrate it into an SAMOEA framework. Experimental results demonstrate that the proposed algorithm significantly outperforms existing SAMOEAs under a strict budget of 50 function evaluations. Moreover, our analysis reveals that the proposed infill criterion improves the sampling efficiency, thereby contributing to the improved performance.

Keywords: Computational Intelligence, Intelligent Systems
Abstract: Geothermal power is a key renewable energy source for achieving carbon neutrality. However, in Japan, its utilization rate has been declining annually, partly due to steam flow interruptions from production wells. To address this, our research group is developing an operational support system using deep learning to detect early signs of such events and provide valve operation guidance. AI models are effective but suffer from low interpretability, which limits system reliability and safety. To overcome this, we integrate a physical model to enhance the trustworthiness of the AI-based system. Specifically, we train a deep learning model on output waveforms from a physical simulator of steam enthalpy at production wells. This allows prediction of steam instability and internal physical parameters. The predicted steam instability matches observed data well, and the estimated physical parameters show high accuracy in most cases. In the future, we aim to build a system that estimates these parameters from real operational data, providing plant operators with insights into subsurface conditions.

Keywords: Computational Intelligence, Intelligent Systems, System Engineering
Abstract: For solving optimization problems, many metaheuristic algorithms have been proposed. Effective algorithms differ from problem to problem, and therefore, employing multiple algorithms can improve performance and reduce the effort required for effective algorithm identification. This paper proposes a novel population method that utilizes multiple algorithms. The proposed method consists of two phases. In the first phase, unlike existing methods, it selects promising algorithms from pre-prepared ones and prioritizes the selected ones. In the second phase, it adaptively adjusts the number of individuals updated using each algorithm. This two-phase approach enables the early identification of promising algorithms and quickly adapts to a problem while maintaining the diversity of the individuals. Numerical experiments demonstrate that the proposed method outperforms existing methods and can appropriately select effective algorithms for various benchmark problems.

Keywords: Computational Intelligence, Intelligent Systems, System Engineering
Abstract: Metaheuristics, including differential evolution (DE), involve several parameters, and their optimization performance depends on the parameter settings. Usually, these parameters are set before applying to problems. In contrast, SHADE using linear population size reduction (L-SHADE) adapts parameters during the optimization process. It has a pair of memories to maintain parameters. Using stored values, new parameters are generated. The memories are updated at each generation based on the history of successful solution updates. However, as the memories are updated, generated parameters become more limited. This limitation may not be appropriate if the fitness landscape is multimodal and complicated. To overcome this limitation, this paper proposes L-SHADE introducing a switch between two pairs of memories based on the non-convex ratio. Numerical experiments demonstrate that the proposed method outperforms an existing method in various benchmark problems.

Keywords: Intelligent Systems, Computational Intelligence
Abstract: The development of games has become more extensive and expensive at an accelerating pace, so making the development more efficient is crucial. In recent years, there has been a lot of research on applying large language models (LLMs), such as GPT, to various fields, and the LLMs have begun to be employed for the development of games. This paper proposes a method for generating game levels using a GPT. In this method, the format of game level data is improved to bridge a gap between natural language processing and game level generation, and the GPT is finetuned using a training dataset in this format. Experimental results demonstrate that the proposed method successfully generates game levels meeting developers' requests.

Keywords: Intelligent Systems, Computational Intelligence
Abstract: In recent years, there have been many attempts to create AI agents for open-ended environments. In particular, research targeting Minecraft, a typical open-ended environment in the field of game AI, has attracted much attention. In the work of L.~Fan et al. MineDojo, which provides a Minecraft simulation environment with an API compatible with OpenAI Gym (now Gymnasium), and MineCLIP, a multimodal model, have been proposed. MineCLIP is trained by using Minecraft-related videos and their audio transcripts from the video-sharing platform YouTube to acquire associations and generate rewards corresponding to prompts from the play screen. It has also been shown that the rewards can be used to train a reinforcement learning agent to solve simple tasks without designing a reward function. In this research, we focus on Minecraft's redstone circuit, which corresponds to a logic circuit, and aim to construct an AI agent that can freely construct circuits like a human player with sufficient knowledge. The specific goal is to construct an AI agent that can accurately recognize the redstone circuit.

Keywords: Transportation Systems, Guidance and Flight Control, Multivariable Control
Abstract: Research and development of autonomous ships focuses on ferries and other vessels equipped with actuators that enable independent berthing/unberthing maneuvers. However, most ships are not capable of maneuvering away from a pier by themselves and require the assistance of tugboats. Existing tugboat control systems have four levels of tug-generated thrust and do not allow for fine-tuning of tug thrust. Therefore, tug control requires predictive control of the ship's motion when tug thrust is applied. This study used a ship maneuvering motion model to pre-calculate the tug control values that predict hull motion. The feedforward control law was used for tugboat control, and the feedback control law was employed for the ship's actuators to cope with model errors and disturbances. The effectiveness of the proposed system was demonstrated by the results of a berthing simulation under wind disturbance.

Keywords: Identification and Estimation, Simulation of Large Systems, Signal and/or Image Processing
Abstract: Estimating material properties in opaque environments is inherently challenging due to complex light scattering, which obscures geometric and reflectance cues. We propose a simulation-augmented framework that integrates multi-angle imaging with synthetic scattering simulations to estimate material-specific optical properties from angular reflection patterns. By capturing reflection cues from multiple fixed viewpoints, our method extracts consistent angular features that are characteristic of each material class. Unlike traditional single-view approaches, we combine physics-based photon trajectory simulation and SimGAN-based refinement with YOLO-based material classification to generate high-fidelity training data, significantly reducing reliance on labeled real-world images. Experiments across twelve material classes demonstrate up to a 24% improvement in mean Average Precision (mAP) compared to single-view baselines, enabled by multi-angle scattering pattern modeling, bounce-based simulation, and SimGAN refinement.

Keywords: Remote Sensing, Identification and Estimation, Signal and/or Image Processing
Abstract: The main goal of this paper is to create a semantic segmentation model that can accurately detect fallen leaves on the pathway. This paper starts by giving a brief introduction to key terms related to machine learning. We cover the basic concepts behind the different types of Machine Learning. The paper then shifts its focus to the different techniques used here to detect the leaves, namely Object Detection and Image Segmentation. We then explain the procedures involved in training the model. After this, we evaluate the model’s performance using different testing images. We also look into other further studies that can be continued in this project.

Keywords: Signal and/or Image Processing, Safety, Environment and Eco-Systems
Abstract: In this paper, we propose novel attention mechanisms in a multi-stream CNN for detecting abnormal behaviors of the driver from an in-vehicle camera. This model inputs the image and the magnitude and direction of the optical flow. The feature maps of the flow magnitude contain significant information about the drivers' behaviors. Hence, we propose two attention mechanisms. One of them is extended CBAM, which consists of the channel and spatial attention, incorporating the feature map of the flow magnitude. The other is a simplified spatial attention, which only uses a 1x1 convolution and global average pooling. The experiment was conducted on the eight drivers to evaluate the performance of the proposed methods by the leave-one-out method. Both extended CBAM and simplified spatial attention are better than the other methods.

Keywords: Computational Intelligence, Signal and/or Image Processing, System Engineering
Abstract: This study evaluates three advanced segmentation models of vanilla Segment Anything Model (SAM), FastSAM, and Semantic-SAM for segmentation of underwater images obtained using a remotely operated vehicle (ROV) to provide detailed and intuitive representations of the fuel debris and surrounding structures. Experimental results demonstrate that FastSAM achieves approximately 50 times faster performance compared to vanilla SAM while maintaining comparable accuracy and exceptional speed. It effectively segments diverse objects, including rectangular boxes and long bars, in images with gradually decreasing water transparency, showcasing its robust segmentation capabilities. However, the lack of an attention framework limits FastSAM's ability to distinguish between remaining regions, resulting in segmentation outputs primarily useful for predefined ranges. These limitations highlight opportunities to refine models like Semantic-SAM, enhancing their adaptability and versatility across diverse applications, including complex underwater environments.

Keywords: Robotic and Automation Systems, Signal and/or Image Processing, Intelligent Systems
Abstract: This paper presents the development of a machine learning-based pedestrian traffic light state detection system that utilizes Retinex color images for sidewalk mobile robots. The system leverages Retinex-enhanced color images to improve visibility and consistency under varying sunlight conditions. To achieve real-time traffic light state detection, we employed the YOLO v8 model in combination with this Retinex-based image preprocessing technique. The Retinex algorithm effectively mitigates the effects of shadows and uneven illumination, enhancing the clarity of traffic light signals. The YOLO v8 model detects the positions of traffic lights from images captured by the sidewalk mobile robot’s camera, while a separate classifier determines the traffic light state as either red or green, enabling the mobile robot to make safe and efficient navigation decisions. Experimental results demonstrate that the proposed method achieves high detection accuracy under various sunlight conditions, outperforming conventional image-based detection techniques. The effectiveness of the proposed system was validated through experiments using a developed sidewalk mobile robot and real-world pedestrian traffic lights.

Keywords: Robotic and Automation Systems, Signal and/or Image Processing, Identification and Estimation
Abstract: This paper describes a novel approach to build a dense three-dimensional (3D) map based on Structure from Motion (SfM) and Multi-View Stereo (MVS) using a general optical camera in extremely dark environments such as the lunar surface and planetary environments. The challenge is to generate a highly visible map that accurately reproduces the geometric structure and color information in a dark environment.To achieve this, we propose a method to generate a geometrically accurate 3D map using the light emitted from a phosphorescent material as a feature point, and then generate a 3D map with improved visibility by pasting a texture image with brightness correction. Experimental results in dark environments demonstrate that our SfM and MVS framework using phosphorescent materials can build a geometrically accurate dense 3D map that includes natural texture information.

Keywords: Robotic and Automation Systems, Intelligent Systems
Abstract: This paper originates from the limitations of traditional Optical Character Recognition (OCR) systems, particularly in recognizing multilingual texts (English and Thai characters) within Automatic License Plate Recognition (ALPR). A common challenge in current OCR models is character confusion, such as misreading between “0” and “O”, and “8” and “B,” which can significantly impact character recognition accuracy. Since OCR performance directly affects the overall effectiveness of ALPR systems, such misclassifications lead to incorrect vehicle identification. Therefore, this paper introduces a rapid prototyping approach for ALPR, focusing on multilingual license plates, especially Thai license plates that include both Latin (English) and non-Latin (Thai) characters. It aims to emphasize the identification and analysis of OCR misclassifications in how characters are recognized by OCR.

Keywords: Signal and/or Image Processing, Identification and Estimation, Mechanical Measurement
Abstract: Food texture is a key factor influencing palatability. This study presents a method for determining the causal relationships between measurement data and sensory evaluation results of food texture. Twelve food samples, categorized into three groups, and ten texture descriptors were selected to analyze changes in food texture during mastication. A measurement system comprising a force sensor, an accelerometer, and a microphone collected data during twelve compression cycles. Sensory evaluation data were also obtained during twelve chewing cycles using the temporary dominance sensations method. The causal relationships between feature values derived from measurement data and texture descriptors from sensory evaluations were determined using the Granger causality test with a vector autoregression model. Most identified causal relationships were found to be reasonable based on our understanding of texture perception and sensory attributes.

Keywords: Robotic and Automation Systems
Abstract: This paper presents the background and development activities associated with participation in AI-Formula, emphasizing the learning outcomes achieved through practical implementation. The team improved the maneuverability of the provided mobility platform by redesigning the caster system with a trail extension and active steering mechanism. In parallel, path-following software was developed utilizing GNSS data, the sliding window method, and YOLOPv2-based perception. Through these activities, the participants acquired practical knowledge and skills in hardware design, software development, and system integration in robotics, while enhancing their problem-solving abilities. The developed hardware and software have been publicly released to support future participating teams.

Keywords: Intelligent Control, Signal and/or Image Processing, Robotic and Automation Systems
Abstract: AI-Formula is a technical challenge planned by SICE and JSAE to acquire knowledge and team problem-solving skills necessary for engineers. In this paper, we describe the algorithm of line recognition and control method for automatic driving control of automobiles, which our research team is conducting in AI-Formula. It also describes the challenges and future plans.

Keywords: Robotic and Automation Systems, Modeling, System Identification and Estimation
Abstract: While the demand for short distance micro-mobility has been recently increasing for the last mile transportation, there is also a growing need for automated driving solutions that enable broader accessibility, allowing people of all ages, including the elderly and children, to move freely. Considering the above, there is a huge potential to evolve the automated driving technology for micro-mobility. In order to advance the reach of autonomous micro-mobility widely at a low cost, we propose a system capable of navigating using only RGB cameras. The image from cameras is utilized to predict the motion of dynamic obstacles by combining deep neural network, homography transformation and Kalman filter. Accurate object localization is important for safe and smooth automated driving, however, it posses unique challenge due to noise in object detection. To tackle this problem, we propose a method to suppress errors in the localization of dynamic objects by considering error propagation in homography transformation. Furthermore, by combining the proposed uncertainty-aware localization framework with decision making algorithm, we achieve reliable path planning at intersections on both sharing sidewalk and roadway in the real world.

Keywords: Robotic and Automation Systems, Autonomous Decentralized Systems, Intelligent Systems
Abstract: In this paper we propose a method to restrict the computational complexity of a speed optimization problem for an autonomous vehicle capable of handling diverse scenarios. The previous speed planning algorithm has succeeded in defining the optimization problem using mixed integer linear programming, where the algorithm implements auxiliary variables at each time frame to represent and adapt to a variety of possible real-world events, such as emergencies and changes in traffic rules, over time. However, introducing a new integer variable per future time frame can increase the solver’s convergence time, hence computational efficiency remains a prominent challenge. In our proposed method, we utilize the optimized trajectory from previous time steps and probabilistically assign the integer variables only to future time frames deemed important. We validate our method by evaluating the reduced time to convergence, and we demonstrate the effectiveness of our method through real-world vehicle experiments.

Keywords: Intelligent Systems, Robotic and Automation Systems, Mechatronics Systems
Abstract: AI-Formula is a technical challenge for student. In the AI-Formula, Mobility drive autonomously through a race course given a mission. Through competing for speed and intelligence in a real-world environment, AI-Formula will provide an opportunity for rising engineers to acquire the skills and technology necessary for next-generation mobility research. This paper details about operations and platform design toward AI-Formula.