Keywords: Control theories
Abstract: This paper investigates the leader-following consensus problem of parabolic multi-agent systems (MASs) with a novel event-triggered mechanism under denial-of-service (DoS) attacks. To address the communication interruption problem caused by malicious attackers launching periodic DoS attacks, we design an event-triggered controller to achieve consensus. Moreover, the allowable frequency and duration of DoS attacks are indicated. The stability of parabolic MASs is analyzed and the Zeno behavior is avoided. Finally, simulation examples effectively validate the theoretical findings.

Keywords: Control theories
Abstract: This paper investigates the secure state estimation problem of semilinear parabolic partial differential equation (PDE) subject to deception attacks. Within the network framework, sampled-data (in time) point measurements of the state are collected from a finite number of sensors distributed in the spatial domain. A secure observer is constructed to cope with randomly occurring deception attacks. Sufficient linear matrix inequality conditions are derived to guarantee the mean-square stability of the estimation error system and find the upper bound on the sampling intervals. Finally, a numerical example is presented to demonstrate the security and robustness of the proposed observer.

Keywords: Control theories
Abstract: This paper establishes a Cayley-Hamilton Theorem for equilateral hypermatrices (EHs) based on the t-product. We introduce a normal-form representation that bijectively reindexes a dth-order EH into an ordinary matrix, and use it to extend the t-product to EHs of general order via a block-circulant embedding. Based on this, we characterize the induced monoid, group, and ring structures, and derive a Cayley–Hamilton Theorem via analytic functions defined through t-power series.

Keywords: Control theories, Nonlinear control and applications
Abstract: Malaria is an infectious disease which causes a global health problem. This paper aims to construct a non-dimensional malaria model and also apply optimal control variables in the form of prevention, treatment, and insecticide. The parameters of the model are fitted to the cumulative number of malaria cases in Indonesia for the period 2015-2025 and parameterized using the least-squares technique. Furthermore, the existence of the optimal control variable in the non- dimensional malaria model is determined through the Pontryagin Maximum Principle. Numerical simulation of the model with the optimal control shows that providing controls in the form of prevention, treatment, and insecticide simultaneously are effective in reducing the number of the infectious human population and also the infectious mosquito population.

Keywords: Control theories
Abstract: This paper investigates the distributed Nash equilibrium (NE) seeking problem for aggregative games in high-order multi-agent systems (MASs) under safety constraints. A distinguishing feature of this work is the proposed hierarchical framework that incorporates a detection-based safety mechanism. By triggering the collision avoidance mechanism within a detection radius, this design eliminates unnecessary global repulsive forces, thereby mitigating the conservatism in traditional barrier function methods. Theoretical analysis indicates that the proposed method guarantees convergence to the NE while ensuring strict satisfaction of the safety constraints. Comparative numerical simulations are provided to verify the effectiveness of the proposed approach, and exhibit advantages over traditional barrier function methods.

Keywords: Control theories
Abstract: This study proposes a self-stabilizing policy iteration (SS-PI) algorithm for the optimal control of complex-valued continuous-time linear systems (CVCTLS). Using mathbb{C}mathbb{R}-calculus, the complex-valued algebraic Riccati equation (CVARE) associated with the optimal control problem is derived without decomposing the CVCTLS into an equivalent real-valued system. Conventional policy iteration (PI) algorithms require an initial stabilizing control policy to solve the CVARE and compute the optimal control policy. To address this limitation, a self-stabilizing policy iteration algorithm is developed. Based on the proposed SS-PI framework, a combined PI scheme is introduced to obtain the optimal control policy without requiring an initial stabilizing control policy. The convergence properties of the algorithm are analyzed. Simulation results demonstrate the effectiveness and computational efficiency of the proposed approach.

Keywords: Control theories
Abstract: In this paper, we investigate neural operator (NO)- based state feedback regulation for a heat equation with spatially varying in-domain coefficients, boundary disturbances, and input delay. The system is subject to distributed and boundary disturbances generated by a finite-dimensional exosystem, and the control objective is to ensure exponential tracking of a reference trajectory. The feedback regulator is constructed via the backstepping method with delay compensation, and the design procedure is significantly accelerated by DeepONet, where DeepONet is employed to approximate the relevant backstepping kernels. Numerical results show that DeepONet achieves nearly two orders of magnitude reduction in computational time compared with conventional PDE solvers. By integrating the learned kernels into the regulator, Lyapunov-based analysis rigorously establishes exponential convergence of the tracking error.

Keywords: Control theories, System identification and modelling, Nonlinear control and applications
Abstract: Gain-scheduled control is an active robust control strategy that adjusts controller parameters based on online measurements of uncertainties, achieving better control performance than a fixed robust controller. However, in some cases, it is not easy to capture the superiority of a gain-scheduled controller by the H-infinity norm of a transfer function, such as an ordinary sensitivity function. Accordingly, this paper proposed a sensitivity analysis of the scheduling parameter.

Keywords: Autonomous vehicles, Deep learning and machine learning, Adaptive systems
Abstract: The advancement of autonomous vehicles (AVs) requires motion planning strategies that not only navigate complex environments, but also account for road surface irregularities. Traditional motion planners often overlook the impact of road damage—such as potholes and cracks—on overall navigational safety. This research proposes a Road Damage-Aware Motion Planning framework that uses Deep Reinforcement Learning (DRL) to optimize driving trajectories in real-time. In this study, a Deep Q-Network (DQN) is employed as the core learning algorithm to discover optimal policies for pothole avoidance. By integrating approximately 9,000 images high-resolution road surface data, the DQN agent is trained to make proactive decisions, such as adjusting speed or performing evasive maneuvers, to mitigate the effects of road distress. The reward function is specifically designed to navigate road irregularities by balancing navigational continuity with proactive pothole avoidance maneuvers. This framework is implemented and validated within the high-fidelity CARLA autonomous driving simulator. The results demonstrate that the trained DRL agent successfully identifies and mitigates the impact of road surface damage, ensuring operational safety during navigation. This study contributes to the development of more robust autonomous systems capable of maintaining high performance even on poorly maintained infrastructure

Keywords: Autonomous vehicles, Deep learning and machine learning, Artificial intelligence
Abstract: Accurate motion forecasting in complex urban environments requires not only capturing agent interactions but also strictly adhering to traffic rules, particularly traffic light signals. Despite the strong performance achieved by state-of-the-art Transformer-based models such as MTR through data-driven learning, they often struggle to explicitly model the physical constraints of traffic signals. This limitation may lead to the “attention dispersion” phenomenon, where critical control signals are diluted by irrelevant inputs, resulting in safety-critical failures such as running red lights. To address this, we propose a prior-informed adaptive weighting mechanism that integrates explicit physical priors into Transformer-based motion prediction. Specifically, a dual-factor design combining spatial proximity and directional relevance is introduced to selectively focus on relevant signals. This approach enforces a soft constraint mimicking human driving logic by focusing on proximal and frontal signals. Experiments on the Waymo Open Motion Dataset demonstrate that the proposed method achieves superior performance across key evaluation criteria, such as mAP and minADE, when compared to the state-of-the-art MTR baseline under identical data settings, while improving rule-consistent trajectory prediction performance.

Keywords: Autonomous vehicles, Deep learning and machine learning, Measurement and instrumentation
Abstract: Inertial navigation system/Global Navigation Satellite System (INS/GNSS) integration is widely used for unmanned aerial vehicle (UAV) navigation, but its performance degrades rapidly when GNSS updates are interrupted. For multirotor platforms, rotor-induced vibration and maneuver-dependent inertial disturbances make the reliability of inertial measurement unit (IMU)-derived pseudo-measurements vary across motion conditions. This paper presents a learning-assisted extended Kalman filter (EKF) fusion strategy for GNSS-denied UAV flight with adaptive confidence scheduling. A gated recurrent unit (GRU)-based temporal encoder summarizes recent IMU motion, and an attention-based probabilistic predictor outputs velocity together with a data-dependent covariance. The predicted mean is used to form pseudo-measurements, while the covariance is mapped to the EKF measurement-noise term to adapt correction strength online. Real-flight experiments on a DJI Matrice 350 RTK platform with a continuous 70 s GNSS outage show that the proposed method yields lower velocity and position errors than long short-term memory EKF (LSTM-EKF) and GRU-EKF baselines, with clearer advantages in turning segments. The results indicate that motion-dependent confidence scheduling improves the robustness of UAV INS/GNSS fusion during GNSS outages.

Keywords: Autonomous vehicles, Intelligent control, Robotics and swarm intelligence
Abstract: Safe coordination at unsignalized intersections remains a critical barrier to realizing connected and cooperative autonomous vehicle deployment, particularly in mixed traffic environments where human-driven vehicles operate without communication protocols and may contest crossing priority unpredictably. We build on our previous work on decentralized signed-network opinion dynamics for all-autonomous intersection coordination, where a conflict-topology communication graph and a commitment-driven belief network jointly drive each autonomous vehicle toward a Go or Yield decision without a central coordinator, facilitating dynamic crossing reordering as traffic conditions evolve. We extend the framework to mixed traffic by explicitly incorporating non-communicating human-driven vehicles as dynamic agents. Human-driven vehicle (HDV) intent can be inferred continuously from onboard sensing via a Backup Control Barrier Function (CBF) evaluated with respect to the vehicle’s distance to the intersection center, which acts as an intent signal that propagates through the signed network to pre-condition autonomous vehicle decisions before physical conflicts arise. Simulation results across two mixed-traffic scenarios demonstrate collision-free coordination and improved throughput performance over a strict First-Come-First-Served (FCFS) baseline under competitive HDV behavior.

Keywords: Autonomous vehicles, Mechatronics, Nonlinear control and applications
Abstract: Coordination among Connected Automated Vehicles (CAVs) is prominent to enable safe and energy-efficient trajectory planning at unsignalized intersections in mixed traffic environments. This can be formulated as a Mixed-Integer Non Linear Programming (MINLP) problem, which is computationally intractable due to its nonlinearity and NP-hardness. In this paper, an approximation method is proposed to retrieve solutions of the MINLP coordination in a computationally efficient way. The method combines the Alternating Direction Method of Multipliers (ADMM) to solve a Mixed-Integer Quadratic Programming (MIQP) approximation of the MINLP problem. The MIQP is built upon sensitivities information, which is then solved via ADMM employing tailored feasibility enforcing functions to improve the quality of approximate solutions. Simulation results show that the sensitivity-based ADMM can obtain feasible and more efficient trajectories than the First-Come, First-Serve (FCFS).

Keywords: Autonomous vehicles, Motion and vibration control, Control theories
Abstract: This paper presents an actuator-aware safety filter using high-order control barrier functions (HOCBF) for safe lane-keeping in all-wheel steer (AWS) vehicles. Unlike conventional methods that ignore actuator delays and risk chattering or safe set violations, we explicitly embed a first-order steering delay model into the barrier constraints. By treating target steering angles as direct control inputs and considering dynamic limits within an extended safe set, our framework ensures physically executable and strictly continuous steering. Simulations demonstrate the filter's ability to proactively adjust commands, preventing lane departures while strictly adhering to hardware constraints.

Keywords: Autonomous vehicles, Motion and vibration control, System identification and modelling
Abstract: Hybrid VTOL unmanned aerial vehicles (UAVs) offer improved operational flexibility by combining vertical takeoff capability with efficient forward flight. However, the presence of flexible wings introduces aeroelastic effects that can significantly influence the aircraft’s dynamics, particularly during the transition phase between hover and forward flight. This paper presents the control design of a flexible-wing tilt-rotor UAV by considering the coupled dynamics between rigid-body motion and structural flexibility. Using a model that represents rigid-flexible body interactions, a forward speed tracking controller is designed using the Linear Quadratic Regulator (LQR) method, along with a full-state observer. Simulation results show that a controller based on rigid-only model was unable to control the growing wing oscillation. While a rigid-flexible controllers can anticipate the flexible response and provide good tracking performance.

Keywords: Autonomous vehicles, Nonlinear control and applications
Abstract: This article addresses the fixed-time trajectory tracking control task with prescribed performance for unmanned surface vehicles (USVs) under input saturation. First, a fixed-time disturbance observer (FTDO) requiring no prior information is constructed to estimate the lumped disturbance consisting of model uncertainties and environmental disturbances. Moreover, an error shifting function is employed to eliminate the stringent requirement for the initial tracking error to lie within the bounds of the prescribed performance function (PPF). Subsequently, a non-singular fixed-time sliding mode controller (NFTSMC) is designed by incorporating the PPF, ensuring both transient and steady-state tracking performance. In addition, an auxiliary variable is incorporated to compensate for actuator saturation. Finally, the fixed-time stability and convergence of the closed-loop system are established through Lyapunov analysis, and numerical simulations are conducted to verify the validity of the proposed approach.

Keywords: Cyber-physical systems and security
Abstract: This article studies the impact of strategic DoS attacks on the linear quadratic Gaussian (LQG) control performance of cyber-physical systems (CPSs), whose goal aims at maximizing the LQG cost at the remote estimator. Different from conventional periodic or Bernoulli-based DoS attack patterns, the concerned one follows a stochastic structure and prioritizes packets carrying essential information, thereby yielding greater damage. On the other hand, the constraint of the average attack rate (AAR) is considered. The relationship between such limit and attack success probability (ASP) is analytically derived. Finally, a numerical example is further conducted to validate the theoretical results.

Keywords: Robotics and swarm intelligence, Control theories, Intelligent control
Abstract: Immersive teleoperation has become a crucial method for collecting human demonstration data to train humanoid robots. However, its adoption is constrained by local computing requirements and limited adaptability to different robot models. To address these challenges, we propose a WebXR-based upper-body bimanual motion retargeting framework that transforms a standard web browser into a lightweight control station. We introduce a logical skeleton abstraction layer that supports user-uploaded models (glTF/GLB) without code modification. We further implement a scale-adaptive analytical inverse kinematics (IK) solver for fixed two-link arm chains, yielding constant-time per-arm computation in the browser runtime. Experiments on the HTC VIVE Focus 3 demonstrate stable 90 Hz rendering, microsecond-level IK computation, controlled mapping linearity over 400 samples (R²=0.999994), and real-operator tracking over 126 VR samples with mean X/Y-axis R²=0.966. By lowering deployment barriers for bimanual teleoperation and data logging, this work facilitates scalable demonstration collection for embodied AI.

Keywords: Cyber-physical systems and security, Nonlinear control and applications
Abstract: Based on a fully actuated system (FAS) method, this study is devoted to a solution to the inference-type human-in-the-loop (HitL) control for cyber-physical systems (CPSs). Combining with the fuzzy inference, a generalized proportional-integral (PI) control is proposed to solve this problem. Concretely, a discrete-time FAS model of the CPSs with human operator is established. Secondly, two assessment indices in relation to environmental risk and human-machine conflict are presented to be used in the fuzzy inference for adjusting the human factor. Then, a generalized PI control by leveraging a FAS method is designed to achieve the expected trajectory tracking with eliminating the original open-loop nonlinearities. A sufficient condition is further constructed to discuss the bounded stability and tracking performance of the closed-loop systems. Finally, an example for trajectory tracking of unmanned aerial vehicle (UAV) is provided to illustrate the feasibility of the proposed method.

Keywords: Artificial intelligence, Deep learning and machine learning, Industrial applications
Abstract: Accurate forecasting of superheated steam temperature is crucial for the safety and efficiency of thermal power plants. However, existing deep learning models typically treat industrial data as generic open-loop sequences, ignoring the inherent feedback mechanisms of the underlying cascade PID control systems, resulting in phase lag and limited performance during transient load changes. To address this, we propose the CASG-former (Control-Aware Spectral Graph Transformer), a novel architecture that bridges physical control logic with data-driven modeling. Specifically, we introduce a Control-State Spectral Modeling module that reconstructs PID error dynamics and analyzes them in the frequency domain to capture global oscillation patterns. Furthermore, an Adaptive Variable Graph Filtering mechanism is designed to disentangle complex multivariate couplings, while a Spectral-Temporal Gated Fusion module dynamically modulates temporal features based on control stability. Extensive experiments on a high-frequency dataset from a real-world thermal power plant demonstrate that CASGformer significantly outperforms state-of-the-art (SOTA) baselines, achieving lower prediction error and effectively minimizing time lag in closed-loop scenarios.

Keywords: System identification and modelling, Artificial intelligence, Industrial applications
Abstract: Developing accurate dynamic models for superheated steam temperature is challenging, particularly when safety regulations restrict open-loop experiments, forcing reliance on closed-loop operational data. Traditional data-driven methods often struggle with closed-loop bias and lack stability guarantees. To address these issues, this paper proposes a Physics- Informed First-Order Neural Linear Parameter-Varying (FONeural- LPV) identification framework. The proposed method integrates deep learning with rigorous control theory through three key contributions: (1) Physical Decoupling, where hard monotonic constraints are embedded into the network to enforce correct gain directionality, effectively isolating plant dynamics from controller logic; (2) Unconditional Stability, utilizing a differentiable Sigmoid-based constraint to strictly bound the inertia pole of the system within the unit interval, ensuring inherent Bounded-Input Bounded-Output (BIBO) stability without complex criteria; and (3) Simulation-Oriented Training, where the model is optimized via sequence-level simulation error minimization to ensure long-term prediction consistency. Experimental validation on a coal-fired power plant demonstrates that the proposed method achieves superior simulation accuracy and physical consistency compared to standard LSTM and unconstrained baselines, offering a structurally simple yet robust digital twin for control applications.

Keywords: Nonlinear control and applications, Intelligent control, Industrial applications
Abstract: This paper investigates robust power tracking control for a single-stage dual-active-bridge (DAB) microinverter operating under parameter uncertainties and external disturbances. By exploiting the intrinsic power transmission characteristics of the DAB topology, the power tracking problem is transformed into an equivalent grid current regulation problem through an averaged small-signal model. A fully actuated system (FAS) model is established to capture the dominant dynamics of the grid current while accounting for modeling errors and parameter variations. Based on this model, a sliding-mode control strategy is developed to achieve finite-time convergence and robust current tracking performance. The proposed controller is designed to suppress both additive and multiplicative uncertainties without relying on precise system parameters. Lyapunov-based analysis is provided to rigorously prove the finite-time stability of the closed-loop system. Simulation results under steady-state operation, leakage inductance mismatch, and grid voltage disturbance demonstrate that the proposed method achieves improved current tracking accuracy and enhanced robustness compared with a conventional proportional-resonant (PR) controller.

Keywords: System identification and modelling, Industrial applications, Nonlinear control and applications
Abstract: Digital Twin systems for electric motor drives require models that support not only accurate forward simulation but also gradient propagation for parameter identification and optimal control. Conventional physical-based simulators treat the numerical integrator as an opaque procedure, blocking gradient flow and preventing their direct use in gradient-based optimization loops. This paper proposes a differentiable physical-operator framework for three-phase induction motors. The continuous-time d-q frame motor model is reformulated as a discrete-time differentiable physical operator via explicit Euler discretization, yielding a mapping that is continuously differentiable with respect to both the system state and physical parameters. Analytical Jacobians are derived in closed form, preserving the bilinear electromechanical coupling structure of the original dynamics. Building on this operator, the discrete adjoint method is applied to compute exact parameter gradients through a backward recursion whose cost is independent of parameter dimension, avoiding the step-size sensitivity and floating-point cancellation that afflict finite-difference alternatives. The resulting framework unifies forward simulation, parameter identification, and trajectory optimization within a single differentiable digital twin, providing a principled foundation for physical-informed gradient-based learning in motor drive systems.

Keywords: Deep learning and machine learning, Intelligent control, Robotics and swarm intelligence
Abstract: Human video demonstrations are an important data source for dual-arm robot manipulation learning. However, structural differences between humans and robots often introduce geometric errors during direct motion mapping and increase the difficulty of policy learning. To address this issue, this paper develops a spatial tracked action reconstruction method with redundant joint action masking for humanoid robot dual-arm cooperative tasks. Spatial tracked action reconstruction is used to explicitly compensate for structural mismatch between human demonstrations and the humanoid dual-arm robot through global orientation correction, keybody mapping reconstruction, local scale adjustment, and jointlevel offset compensation. A redundant joint action masking mechanism is further introduced to compress the highdimensional robot action space into a task-relevant action subspace, thereby reducing the search dimension and training difficulty of the reinforcement learning policy and enabling stable trajectory tracking control through deep reinforcement learning. Simulation results of dual-arm cooperative tasks show that the method reduces z-axis position errors of key bodies by approximately 71% and 3D spatial errors at keyframes by approximately 74%; reinforcement learning policies trained on the reconstructed trajectories achieve faster convergence and higher tracking accuracy than the baseline method.

Keywords: Control system education, Control theories, Industrial applications
Abstract: With the deepening research into Networked Control Systems (NCS) and Multi-Agent Systems (MAS), a significant gap persists between theoretical simulation and physical practice, posing a severe challenge in education and research. This paper presents a novel pedagogical practice framework implemented within the NCSLab remote laboratory to bridge this gap. Our core contribution is a Unified Modeling Paradigm: we extend the functionality of the graphical Subsystem module, allowing it to serve as a ``Simulation Unit" by default, or transform into a ``Physical Device Proxy" when assigned an IP address. This enables a ``Model-to-Deployment" workflow, where a single graphical model supports two paths: (1) a Simulation path, where the platform compiles the monolithic model for server-side numerical verification; and (2) a Compile path, in which triggers an automated backend pipeline that Auto-Splits the model, concurrently compiles N separate executables, and performs a ``one-click deployment" to multiple distributed hardware controllers. This approach abstracts the underlying engineering complexity, allowing scholars to focus on algorithm design. The framework's efficacy is validated with a distributed fan control experiment.

Keywords: System identification and modelling, Control system education, Measurement and instrumentation
Abstract: Existing online control experiment platforms are constrained by insufficient real-time performance and timing determinism in the software-based execution of control modules on general-purpose processors. To address this issue, this paper presents a Field Programmable Gate Array (FPGA)-based hardware implementation method for fundamental control modules. First-order and second-order lag elements are selected as representative cases. Continuous-time models are transformed into FPGA-oriented recursive hardware structures through forward Euler discretization and fixed-point representation. A complete verification link is established using Verilog implementation, a Universal Asynchronous Receiver/Transmitter communication module, and Python-based host-side visualization. The proposed modules are verified through functional simulation and FPGA board-level tests. Post-synthesis results show that both implementations satisfy the target clock constraint of 50 MHz, with estimated maximum operating frequencies of 186.6 MHz and 116.7 MHz for the first-order and second-order cases, respectively, while requiring only 68/170 Slice Look-Up Tables (Slice LUTs) and 62/64 Slice Registers. These results demonstrate that representative fundamental control modules can be implemented on FPGA with good timing feasibility and low resource cost, providing a practical hardware execution path for extending online control experiment platforms from software-side execution toward hardware-side deployment.

Keywords: Adaptive systems, Control theories, Nonlinear control and applications
Abstract: This paper addresses the problem of distributed event-triggered (ET) control for multi-vehicle platoon under unreliable network communication and input saturation. Unmeasurable states and disturbances are approximated via adaptive extended state observers (AESOs). Moreover, to deal with unreliable topology connections and limited bandwidth resource in vehicle-tovehicle (V2V) networks, an event-based fully distributed estimator is designed to generate the smooth differentiable estimations of virtual leader’s system matrix and states. This ET mechanism relieves pressure of network bandwidth. Combining H1 control method and Lyapunov stability theory, an anti-interference robust platoon control strategy is proposed. It ensures the boundedness of all signals and the stability of the closedloop system. Finally, simulation studies are shown to verify the effectiveness of this distributed ET control strategy.

Keywords: Energy Systems, Industrial applications
Abstract: Model predictive control (MPC) using Light Detection and Ranging (LIDAR) measurements can improve power regulation and reduce structural loads in large-scale wind turbines. However, in practical implementations, the effective LIDAR wind preview time is often limited by measurement quality and data availability, which may degrade control performance under realistic operating conditions. This work proposes a multi-input, multi-output (MIMO) MPC framework using LIDAR wind measurements, in which the LIDAR preview time is decoupled from the MPC prediction horizon. The proposed approach is compared with a single-input, multi-output (SIMO) MPC design through simulation studies on a 5 MW linearised wind turbine model under two turbulent wind conditions. Ideal LIDAR measurements without noise or distortion are assumed, while the preview time is constrained to emulate realistic operational limitations. Results show that the proposed MIMO MPC improves both power regulation and load reduction, particularly under limited LIDAR preview time and high wind speed conditions.

Keywords: Autonomous vehicles, Control theories
Abstract: Vehicle platooning control improves traffic efficiency, safety, and energy use by coordinating connected vehicles' longitudinal motion. In this paper, a decoded information (DI)-based integral sliding mode (ISM) control scheme is proposed to tackle platooning control challenges of vehicle platoon systems subjected to deception attacks. First, an event-triggered encoding-decoding (ETED) mechanism is designed based on an observer affected by deception attacks to reduce communication burden and enhance data security. Then, by constructing a DI-based ISM surface, equivalent and actual ISM controllers are designed to simultaneously achieve desired platooning performance and guarantee the finite-time reachability of the specified ISM surface. Furthermore, sufficient conditions for determining the gains matrices of the ISM controller and the observer are obtained through the feasibility solution of a set of LMIs. Ultimately, the efficacy of the designed control law is validated via MATLAB simulations.

Keywords: Cyber-physical systems and security
Abstract: A novel and explicit anomaly propagation model is proposed for networked multi-agent systems to describe how local faults or attacks spread through directed interaction networks. The model characterizes the evolution of anomaly intensity at each node by combining local decay effects with propagation terms induced by multi-hop neighbor interactions. To avoid trivial averaging of anomaly influence, threshold-based nonlinear couplings are incorporated into the propagation dynamics. It is demonstrated that the resulting anomaly evolution can be analyzed as a well-posed, positive dynamical system, which enables the derivation of sufficient conditions for suppressing transient anomalies. For persistent anomaly input, the existence and boundedness of steady-state anomaly profiles are established. Moreover, under mild assumptions, the steady-state anomaly intensity is shown to decay exponentially with the graph distance from the anomaly sources. These results clarify how network topology and propagation parameters influence the spatial distribution of anomalies in multi-agent systems.

Keywords: Control theories
Abstract: This paper investigates maximum power point tracking (MPPT) for marine current turbine (MCT) generation units subject to network-induced measurement delays. A neural network (NN) is developed as a software sensor to reconstruct the effective flow velocity using delayed electrical variables and rotor speed, thereby removing the reliance on fragile in-situ flow sensors. Based on the estimated velocity, an adaptive P controller is designed to track the optimal rotor speed reference and enhance energy capture under nonlinear electromechanical coupling. Simulations demonstrate improved tracking accuracy, reduced oscillations in the presence of time-varying delays

Keywords: Control theories, Nonlinear control and applications
Abstract: This paper investigates global finite-time stabilization of second-order systems under asymmetric input saturation. A nonlinear controller is constructed using smooth bounded functions, designed to avoid hard switching and enable explicit settling time estimation. Finite-time stability is established via Lyapunov analysis and homogeneity theory, with an explicit upper bound for the settling time. The method is further extended to disturbance-rejection scenarios using a fixed-time disturbance observer. Simulations validate the effectiveness and robustness of the proposed approach.

Keywords: Artificial intelligence, Autonomous vehicles, Cyber-physical systems and security
Abstract: Federated learning has garnered extensive attention in the realm of Connected and Autonomous Vehicles for its ability to coordinate multi-vehicle collaborative training while preserving data privacy. Given the communication uncertainties and data heterogeneity inherent in moving vehicles, prototype-based personalized federated learning emerges as a practical paradigm for deployment. However, under Non-IID scenarios, existing methods often suffer from performance degradation due to coupled optimization. The joint training of the feature extractor and classifier on clients can induce optimization conflicts, and the passive average aggregation of local prototypes on the server leads to prototype collapse. To address these issues, we propose FedDPC, a Federated Dual-Decoupled Prototype Calibration framework. A Global Prototype Learner is introduced on server with Orthogonal Prototype Contrastive Learning to construct representative and orthogonal global prototypes. And a Top-Down Dual Calibration strategy is designed on clients to decouple the optimization process of the feature extractor and classifier. By leveraging Virtual Prototype Replay to calibrate the classifier, FedDPC effectively mitigates the optimization conflicts between the biased classifier and global prototypes. Extensive experiments demonstrate that FedDPC outperforms existing methods in various data heterogeneity scenarios.

Keywords: Control theories
Abstract: This paper investigates finite-/fixed-time consensus of multi-agent systems (MASs) with higher-order interactions. Different from existing results that mainly rely on pairwise couplings, a unified interaction framework is established by integrating pairwise and higher-order couplings into a hybrid generalized Laplacian representation. Under this framework, two mixed-power protocols with similar forms but different consensus effects are studied. Specifically, an edge-wise protocol is proposed to achieve average consensus, while an operator-based protocol is developed to achieve complete consensus. On this basis, sufficient conditions for the corresponding finite-time consensus are further derived. Numerical simulations are provided to verify the effectiveness of both protocols. Moreover, the simulations not only compare the behaviors of the two protocols, but also examine the effects of higher-order interactions and controller parameters on the consensus process.

Keywords: Control theories, Autonomous vehicles
Abstract: This paper investigates the distributed state estimation problem for networked systems by utilizing both quantized absolute measurements and relative measurements. To address the quantization errors in the absolute measurements, an ellipsoidal outer-bounding method is first employed to characterize the corresponding quantization error set. A distributed set-membership estimation method that aggregates absolute and relative measurement information is then designed to obtain the local state estimation ellipsoid for each node. Subsequently, an ellipsoidal fusion strategy is introduced based on the local estimation results to improve the overall state estimation performance of the considered system. Finally, simulation results involving an automated guided vehicle verify the effectiveness and superiority of the proposed method.

Keywords: Communication, Autonomous vehicles, Robotics and swarm intelligence
Abstract: Monitoring riverine environments requires collecting spatially comprehensive data across complex and often inaccessible channel networks. This paper considers the challenge in deploying diverse sensors on heterogeneous mobile platforms to collectively capture high-resolution data across these dynamic river systems. In particular, increasing pressures such as land-use change, hydromorphological alteration, and climate-driven shifts in flow regimes highlight the growing need for detailed spatial data on river networks to support biodiversity protection and sustainable water management. The growing use of heterogeneous riverine remote sensing, such as sonar for bathymetric surveys, above-water LiDAR for river bank geomorphology, aerial imagery for channel mapping, creates demand for scalable field operations. With the need to coordinate the data gathering and to provide robust links for high bandwidth sensor streams, we identify that multi-robot system and Wi-Fi mesh networking are a promising scalable solution to improve remote monitoring quality. Furthermore, we present reviews and comparison of the existing Wi-Fi mesh technologies, analysing the integration of hardware layers, routing protocols, and middleware, together with communication-aware control and navigation algorithms that enable real-time coordination for the mobile platforms. Thus, providing a reference for implementing a robust mesh communication solution for complex riverine monitoring missions.

Keywords: Cyber-physical systems and security, Deep learning and machine learning, Industrial applications
Abstract: In modern cyber-physical systems (CPSs), state estimation is increasingly threatened by intelligent intermittent Denial-of-Service (DoS) attacks. Unlike traditional attacks, intermittent DoS attackers with learning capabilities can dynamically infer deterministic defense patterns and preemptively block critical sensor data, which greatly threatens the accuracy and stability of estimations. To address this challenge, we formulate the resilient observation scheduling problem as a two-player, zero-sum Markov game between an adaptive defender and a smart attacker. We propose a novel game-theoretic Deep Reinforcement Learning (DRL) framework, featuring a customized Proximal Policy Optimization (PPO) algorithm integrated with a curriculum-guided Fictitious Self-Play (FSP) mechanism. This approach effectively overcomes the non-stationarity and "strategy cycling" pathologies inherent in multi-agent adversarial training. Furthermore, we propose an emergent, physics-inspired "decoy" mechanism: guided by the stability threshold of system control theory, the defender autonomously learns to randomly sacrifice low-value data packets to absorb interference energy, thereby transmitting critical data packets more securely. Simulations demonstrate that our proposed framework successfully converges to a robust mixed-strategy Nash Equilibrium, significantly outperforming conventional fixed-order strategies and ensuring long-term system stability under extreme adversarial conditions.

Keywords: Control theories, Communication, System identification and modelling
Abstract: High bandwidth-delay communication networks would require high-performance transport protocols. The Rate Control Protocol (RCP), based on explicit rate feedback, can potentially be a high-performance transport protocol. RCP estimates the transmission rate using feedback from queues. There are two design considerations for the feedback: (i) a combination of rate mismatch and queue size, and (ii) only rate mismatch. Both design considerations ensure a fair allocation of rates among users. In this paper, we derive sufficient conditions for the local stability of a proportionally fair RCP, under both design considerations, across network topologies ranging from a single bottleneck to multiple bottleneck links. We focus on a network operating with small buffers, assuming heterogeneous round-trip delays for users. The stability conditions are decentralized and scalable as the number of bottleneck links increases from one to many. Importantly, the stability conditions do not depend on the number of users, the link capacity, or the round-trip delays of the users. Finally, we validate the stability conditions using numerical simulations for a single bottleneck topology.

Keywords: Cyber-physical systems and security
Abstract: Industrial control systems (ICS) are increasingly targeted by cyber attacks that can undermine physical processes and disrupt critical infrastructure. Nevertheless, detecting and interpreting such attacks remain a challenge as many intrusion detection methods reduce system deviations to scalar anomaly scores. While such approaches are effective for binary detection, they discard directional information that captures the structure of attack behavior. To address this issue, this paper proposes a Subspace Deviation Encoding (SDE), where normal subspace is characterized using principal component analysis (PCA), and the deviations representing the cyber-physical attacks are retained as full residual vectors. Experiments on a gas pipeline control dataset demonstrate that the resulting deviation manifold occupies a low-dimensional region within the residual space and remains highly stable (≈0.999 mean cosine similarity) under bootstrap analysis and temporal split evaluation. Furthermore, compared to scalar residual magnitude achieving 46.9% attack classification accuracy, the proposed 4D directional encoding achieves an accuracy of 76.3%, corresponding to a 62.7% relative improvement. These findings indicate that the directional deviation information preserves structural properties of cyber-physical attacks that are lost in traditional anomaly scores.

Keywords: Cyber-physical systems and security, Autonomous vehicles, Nonlinear control and applications
Abstract: This paper proposes a sliding mode observer (SMO) for the simultaneous estimation of sensor and actuator attacks in nonlinear systems. The proposed observer extends the classical Luenberger observer by incorporating a switching term and a nonlinear injection mechanism. The observer gain matrices are obtained through linear matrix inequalities (LMIs). The effectiveness of the observer is demonstrated on a three-wheeled omnidirectional robot model, where a sensor-actuator attack scenario is simulated. The results demonstrate that the proposed SMO accurately reconstructs both sensor and actuator attack signals, highlighting its efficacy and potential for enhancing the resilience of autonomous systems.

Keywords: Cyber-physical systems and security, Deep learning and machine learning, Industrial applications
Abstract: Predictive state recognition struggles with heterogeneous data and class imbalance, often confusing postmaintenance recovery with early faults. Furthermore, the opacity of deep models hinders the physical interpretation of fault propagation dynamics. To elucidate the spatio-temporal propagation mechanism of equipment faults and provide physical interpretability of fault evolution, this work proposes a novel framework integrating gated residual adaptive feature attention (G-RAFA) and class-adaptive dual-constraint learning (CADCL). Specifically, the G-RAFA module is designed to address the interference of high-frequency noise, utilizing a learnable gating mechanism to preserve critical fault precursor information. Complementarily, the CA-DCL strategy is employed to overcome severe class imbalance and decision boundary ambiguity, synergizing refined focal loss and center loss to enhance the model’s discriminative capability. Experimental results on the Microsoft Azure PdM dataset demonstrate that the proposed method significantly outperforms state-of-the-art algorithms in terms of robustness and effectiveness.

Keywords: Communication, Control devices, sensors and actuators, Motion and vibration control
Abstract: Accurate acoustic source localization in reverberant environments remains a fundamental challenge for applications ranging from video conferencing to industrial monitoring and wireless acoustic sensor networks. While the Deconvolution Approach for the Mapping of Acoustic Sources (DAMAS) achieves excellent static source localization, its performance for dynamic source tracking—a core requirement for real-world deployment—is fundamentally limited by unaddressed challenges such as time-varying transfer functions and motion-induced artifacts. This paper presents an enhanced DAMAS algorithm that bridges this gap through three key modifications designed to address motion artifacts: temporal windowing (10 ms) to isolate direct paths, frequency-weighted cross-spectral matrices for improved resolution, and adaptive relaxation (ω = 0.1 − 1.8) for faster convergence. Experimental evaluation demonstrates that the proposed enhanced DAMAS algorithm achieves 0.04 m error for static sources (a 33% improvement over standard DAMAS) and maintains 0.16 m average accuracy for dynamic tracking along a 2.12 m trajectory—representing a 53% improvement over the baseline dynamic results.

Index Terms DAMAS, acoustic source localization, microphones arrays, dynamic tracking, beamforming

Keywords: Deep learning and machine learning
Abstract: Decentralized Federated Learning (DFL) serves as a distributed learning framework that enables collaborative model training while preserving privacy, but its convergence rate is still constrained by factors such as data heterogeneity and low-connectivity topologies. Motivated by these observations, we propose two adaptive optimization-based DFL algorithms: DFedAdam and DFedAdamM. DFedAdam adopts the Adam algorithm as its local optimizer to mitigate data heterogeneity via adaptive learning rates. DFedAdamM further enhances this approach by integrating a decaying momentum term during aggregation. The convergence analysis demonstrates that DFedAdam attains a convergence rate of O(1/T ) for non-convex objectives. Evaluations conducted on MNIST and CIFAR-10 indicate that the proposed algorithms consistently surpass DFedAvgM, DFedSAM, and other baseline methods across both IID and non-IID settings, with more pronounced improvements in sparsely connected scenarios. Our methods provide effective insights for accelerating DFL convergence, suitable for heterogeneous data or sparse network connectivity.

Keywords: Health systems, Measurement and instrumentation, System identification and modelling
Abstract: Conventional bed-exit monitoring systems often suffer from false alarms, signal drift, privacy concerns, or costly installation. This paper presents a different sensing strategy for bed-exit monitoring: instead of suppressing the influence of the human body on a UHF RFID reader antenna, the proposed design exploits that perturbation as the sensing signal itself. The system uses a flexible S-shaped RG178 transmission line with a total physical length of 2.283 m and a total electrical length of 10λg at 920 MHz, together with a front-end monitoring tag whose backscattered received signal strength indicator (RSSI) acts as a distributed probe of the entire structure. Human contact along the line produces a clear RSSI drop because body loading perturbs the standing-wave distribution and shifts the resonant condition. Full-wave simulation and prototype experiments show that the antenna reaches S11 = −24.2 dB at 920 MHz, maintains stable operation under bending and body contact, and yields a larger RSSI reduction for human contact (6–9 dB) than for common inanimate objects (2–5 dB). Using normalized RSSI derived features with a radial-basis-function support vector machine, the proposed system achieves 200/200 correct decisions in binary in-bed versus out-of-bed trials, corresponding to an observed accuracy of 100% with a 95% confidence interval of [96.4%, 100%]. The design offers a low-cost and privacy preserving route toward flexible bed-exit monitoring in smart care envi

Keywords:

Keywords: Control theories, Energy Systems
Abstract: In this paper, we address the problem of power sharing in buck converter networks. Utilizing the Krasovskii passivity property of the networks, we propose a distributed output feedback controller that achieves power sharing under unknown constant load demand. The effectiveness of the proposed controller is demonstrated through numerical simulation.

Keywords: Motion and vibration control, Control theories
Abstract: This paper studies attitude tracking and vibration control of three-dimensional (3D) flexible manipulators subject to input amplitude and time-varying output constraints. To capture the coupling between the rigid base motion and flexible link deformation, an infinite-dimensional model described by coupled partial differential equations (PDEs) and ordinary differential equations (ODEs) is derived via Hamilton's principle, with modified Rodrigues parameters used for attitude representation. An input-output constrained boundary control scheme is developed using the backstepping method without model reduction, where the amplitude constraints of control inputs are handled by incorporating hyperbolic tangent functions and auxiliary systems. Meanwhile, an asymmetric barrier Lyapunov function (BLF) is introduced in the control design process to ensure that the attitude tracking error always remains within the specified time-varying constraints. The Lyapunov theory-based analysis demonstrates the boundedness of all closed-loop signals and the exponential convergence of both the attitude tracking error and elastic deformations. Simulation results validate the effectiveness of the proposed control scheme.

Keywords: Control theories, Nonlinear control and applications
Abstract: The continuous-discrete projection filter offers a rigorous framework to approximate the solution of the nonlinear state estimation problems. However, it suffers from numerical instability during the prediction phase when integration errors force the natural parameters outside their admissible domain. To address this issue, we introduce the sum-of-squares (SOS) relaxation to constrain the evolution of the natural parameters within the admissible domain.By parameterizing the underlying SOS matrix using the log-Cholesky map, we derive a projected ordinary differential equation (ODE) that inherently preserves the necessary positivity constraints without requiring the computationally expensive online optimization checks associated with previous semi-infinite programming approach. We provide a theoretical derivation of this positivity-preserving propagation scheme and present the explicit SOS-relaxed evolution equations for the Gaussian case.

Keywords: Autonomous vehicles, Motion and vibration control, Nonlinear control and applications
Abstract: TThis paper addresses the path-following problem of unmanned aerial vehicles under bounded control inputs. The proposed guidance strategy employs nested saturation functions to confine the control inputs within a predefined set, enabling the UAV to converge to the desired path regardless of its initial engagement geometry. The proposed strategy remains applicable to any generic smooth path. The effectiveness of the proposed controller is demonstrated through simulations across various scenarios, including straight-line, circular, and sinusoidal paths.

Keywords: Motion and vibration control, Adaptive systems, Control theories
Abstract: Attitude control of flexible spacecraft faces significant challenges due to multi-source heterogeneous disturbances. These disturbances appear in flexible appendage vibrations, inertia variations, and actuator errors, resulting in intertwined interactions that degrade pointing accuracy. This article proposes a composite feedback fault-tolerant attitude control approach that synergizes refined adaptive disturbance observer and robust adaptive fault estimation observer methodologies, enabling disturbance separation and targeted compensation of multi-source disturbances. A deep-coupled attitude dynamics model is first established to characterize the interplay between vibrations, uncertainties, disturbances, and actuator faults. A novel refined adaptive disturbance observer is designed to accurately estimate vibrations by leveraging prior structural knowledge while attenuating external disturbances. Subsequently, a H_infty based robust adaptive fault estimation observer is employed to estimate lumped inertial uncertainties and actuator faults. Finally, the fault-tolerant control law drives the errors to achieve the desired attitude while compensating for estimated lumped faults, rejecting vibrations, and residual disturbances. This unified architecture decouples disturbance by isolating and minimizing overcompensation. Numerical simulations demonstrate the efficacy of the proposed control and improved attitude stabilization of flexible spacecraft under multiple disturbances.

Keywords: Control theories
Abstract: Controllability scores provide control-theoretic centrality measures that quantify the relative importance of state nodes in networked dynamical systems. We establish a structural connection between finite-time controllability scoring and approximate optimal experimental design (OED): the finite-time controllability Gramian decomposes additively across nodes, yielding an affine matrix model of the same form as the information-matrix model in OED. This yields a direct correspondence between the volumetric controllability score (VCS) and D-optimality, and between the average energy controllability score (AECS) and A-optimality, implying that the classical D/A invariance gap has a direct analogue in controllability scoring. By contrast, we point out that controllability scoring generically admits a unique optimizer, unlike approximate-OED formulations. Finally, we uncover a long-horizon phenomenon with no OED counterpart: source-like state nodes without a negative self-loop can be increasingly downweighted by AECS as the horizon grows. Two numerical examples corroborate this long-horizon downweighting behavior.

Keywords: System identification and modelling, Deep learning and machine learning, Artificial intelligence
Abstract: We introduce the Long-Short-Range Neural Network (LSR-Net), a novel neural operator architecture designed for data-driven forward evolution modeling, and extends it to the prediction of nonlinear fluid dynamics. LSR-Net learns the evolution operator of a dynamical system solely from pairs of initial and future state snapshots, which splits the learnable integral kernel into long-range (LR) and short-range (SR) components within stacked network blocks. While the SR component uses standard convolutions to capture local dynamics, the LR component employs a sum-of-exponentials (SOE) representation. This allows for the efficient computation of global interactions as a trainable Fourier multiplier, reducing computational complexity to mathcal O(n log n) where n is the number of pixels in an input snapshot and requiring only a few parameters per channel. LSR-Net is evaluated on three challenging 2D benchmarks: the coupled Burgers equation, the wave equation with a spatially varying coefficient, and the nonlinear shallow water equation {(SWE)}. Results demonstrate that LSR-Net significantly outperforms a baseline short-range network (SR-Net) in predictive accuracy, achieving substantially lower relative errors by effectively capturing both local fine-scale structures and crucial global pattern interactions.

Keywords: Control theories
Abstract: This paper presents a novel backstepping-based approach for the prescribed-time deployment of multi-agent systems subject to semi-Markov switching topologies in 2D/3D spaces. By leveraging a distributed communication protocol and continuum approximation, the collective dynamics of the agents are modeled as a linear Korteweg–de Vries equation. In particular, the topological weights are characterized as semi-Markov switching processes to better reflect realistic communication environments. We further employ the backstepping method to construct a target system with a time-varying coefficient and design a boundary feedback controller, which transforms the original system into the target system. The prescribed-time stability of the target system is theoretically established. Finally, numerical simulations are conducted to verify the effectiveness of the proposed approach.

Keywords: Control theories
Abstract: This work studies the distributed optimal consensus control problem involving a class of uncertain linear multi-agent systems. Given a desired upper bound on a global quadratic cost functional, we design control laws for homogeneous multi-agent systems with uncertainties that ensure the closed-loop system reaches consensus and the cost remains below the upper bound. First, we show the conditions a given feedback gain needs to satisfy to achieve consensus and also be suboptimal for a given upper bound for all uncertainties in a predefined uncertainty set. Next, a method for calculating such feedback gains based on a Riccati inequality is presented. A simulation illustrating the results obtained is presented.

Keywords: Control theories
Abstract: Distributed state estimation is critical for applications such as surveillance, autonomous navigation, and wide-area monitoring, where sensor agents must cooperatively track targets using only local measurements and neighbor-to-neighbor communication. Existing distributed filters have been shown to achieve accurate estimation even under sparse inter-agent communication and limited sensing ranges. However, many of these methods rely on consensus parameters that depend on global properties of the communication graph, such as the maximum degree of the graph, and are therefore sensitive to changes in network topology. This limitation is particularly significant in sensor networks with mobile agents, where communication links change over time. This paper presents a Dynamic Generalized Kalman Consensus Filter for target tracking in sensor networks with switching communication topologies. The proposed algorithm computes information-based consensus weights using only locally available quantities, eliminating the need for global network parameters. Numerical simulations demonstrate that the proposed algorithm maintains estimation accuracy under switching network topologies and outperforms existing distributed filters in the considered tracking scenario.

Keywords: Control theories
Abstract: This paper studies a Nash equilibrium seeking problem of linear multivariable multi-agent systems. Under the strict relative degree condition, an output feedback based algorithm is constructed that drives the output of each player to the Nash equilibrium asymptotically, even when the system matrices are uncertain. Our algorithm is composed of an auxiliary NE seeking dynamics and the output feedback tracking controller, where the former provides an estimation of the NE and the latter assures the convergence of the output strategies of all players to the estimation of the Nash equilibrium. In contrast with existing results, a reduced-order output feedback tracking controller is incorporated in our algorithm, leading to a lower dimension of the algorithm.

Keywords: Control theories, Adaptive systems, Nonlinear control and applications
Abstract: Memory-based adaptive control methods parameter convergence under finite excitation conditions. However, for systems with piecewise-constant uncertain parameters, stored data may retain previous parameter information after parameter changes, leading to less accurate parameter-estimation performance. Existing reset-based methods perform memory resets at predetermined instants, which may be ineffective when the parameter changes do not coincide with the selected reset times. This paper proposes a residual-based reset mechanism for adaptive systems based on regression filtering and Dynamic Regression and Extension Mixing (DREM) with piecewise-constant parametric uncertainties. The proposed method detects parameter variations directly from the filtered residual signal and activates reset actions near the actual parameter-jump instants. The simulation results demonstrate improved parameter-estimation performance and tracking accuracy compared to the existing reset-based schemes in the literature.

Keywords: Control theories, Intelligent control, System identification and modelling
Abstract: The fuel economy is significantly influenced by the energy management strategy (EMS) in parallel hybrid electric vehicles. This paper proposes a two-time-scale adaptive equivalent consumption minimization strategy (ECMS) based on extremum seeking. In the fast time scale, an exponential state of charge (SOC) feedback mechanism is introduced to rapidly correct charge deviation, while in the slow time scale, an extremumseeking algorithm is employed to optimize the base value of the equivalence factor online, thereby achieving a balance between charge-sustaining performance and long-term fuel economy. To evaluate the proposed method, simulations are conducted on a P2 parallel hybrid electric vehicle against the SOC-feedback-based ECMS and the extremum-seeking-based ECMS (ES-ECMS) .The results demonstrate that, compared with the other strategy, the proposed strategy can effectively improve fuel economy.

Keywords: Control theories, Autonomous vehicles, Intelligent control
Abstract: To address the problem of external disturbances faced by quadrotor Unmanned Aerial Vehicles (UAVs) in intricate flight conditions, this research presents a neural network-based anti-disturbance control strategy. The control architecture adopts a dual closed-loop design. Regarding trajectory regulation, a hybrid adaptive architecture merging backstepping theory with neural learning is synthesized to suppress stochastic uncertainties and gust-induced oscillations. Utilizing the approximation capability of Neural Networks, this framework achieves instantaneous identification and offsetting of total external interferences,which refines the translational tracking performance. A sliding mode controller (SMC) is designed for the attitude controller. This design addresses internal parameter perturbations and satisfies rapid response requirements. Consequently, it ensures that orientation states precisely track the reference trajectories from the translational loop while enhancing structural resilience. By employing the Lyapunov stability criterion, the asymptotic convergence and equilibrium of the entire feedback framework are rigorously validated. Traditional PID regulators and non-hybrid structures show limitations compared to our suggested approach. Simulation results indicate that our hybrid strategy enhances tracking accuracy and stabilizes much quicker. Notably, its ability to reject disturbances remains exceptional under high-interference conditions.

Keywords: Control theories, System identification and modelling
Abstract: This paper highlights the comparison, in terms of advantages and drawbacks, between sliding mode differentiators and sliding mode observers. Beyond the case where the signal to be derived comes from an unknown system, for which a differentiator is obviously required, the choice between a differentiator and an observer is not so simple. This is even more true considering that, in information theory, any unused information is lost information, e.g. time scale, structural properties... This issue has been addressed indirectly in Kalman filtering, where both the degree of confidence in the model and the quality of the signal are taken into account. However, this approach still relies on the availability of a model and on some assumptions of linearity. In this note, we focus on nonlinear settings and discuss several scenarios illustrating how differentiators and observers can be combined or selected to optimally exploit available information.

Keywords: Adaptive systems, Control theories
Abstract: We propose two noise-aware policy iteration methods for nonlinear systems with both process and measurement noise. First, assuming exact Koopman linearization is achievable via a finite-dimensional set of observables, we develop a data-driven, Koopman-lifted method that optimizes a quadratic control objective while mitigating the impact of noise on learning, and we provide quantitative stability and performance guarantees. Second, we present a kernel-based variant that executes the former method directly in the data space. This variant has computational cost that scales with the number of samples and enables efficient tuning of controller hyper-parameters, making it suitable for complex systems that require rich (possibly infinite-dimensional) observable spaces. The effectiveness of the proposed kernel-based method is demonstrated on a noisy Duffing oscillator.

Keywords: Adaptive systems, Control theories, Deep learning and machine learning
Abstract: This study investigates the application of an adaptive linear quadratic (LQ) anti-sway control method based on Q-learning for an underactuated rotary pendulum system. The proposed approach formulates the optimal control problem using a quadratic Q-function and updates the state feedback gain directly from input–output data through a gradient descent method, without requiring explicit knowledge of the system dynamics. The learning update is implemented with a prescribed learning rate to regulate the convergence speed and adaptation stability. In addition, a decaying exploration signal is incorporated into the control input to ensure sufficient excitation during the learning process while gradually reducing the influence on closed-loop performance. The simulation results show convergence of the state feedback gain and effective suppression of pendulum sway. These results demonstrate the feasibility of the proposed method for underactuated mechanical systems and serve as a preliminary validation step toward future nonlinear and experimental implementation.

Keywords: Adaptive systems, Control theories, Intelligent control
Abstract: Proportional–integral–derivative (PID) control systems are widely employed in industrial processes because of their simple structure and intuitive physical interpretation of parameters. In particular, tuning methods for systems with nonlinear or varying characteristics have been actively investigated. Against this background, adaptive and learning-based methods have been increasingly required to satisfy the plant requirements, including two-degree-of-freedom (2DOF) PID control. In this paper, a 2DOF control system is newly proposed that adjusts PID+FF control parameters based on plant operational data using a data-driven approach, specifically a database-driven control method. Its effectiveness is quantitatively verified through numerical examples.

Keywords: Adaptive systems, Control theories, Nonlinear control and applications
Abstract: In this study, we propose an asymptotic consensus tracking control strategy that minimizes communication requirements by reducing the volume of transmitted signals and employing a quantized communication mechanism. We demonstrate that the designed method not only achieves asymptotic consensus but also guarantees higher-order tracking properties, ensuring that all signals within the closed-loop system remain bounded. Finally, we present a simulation with quantized communication to validate the effectiveness of the proposed scheme.

Keywords: Adaptive systems, Control theories, Nonlinear control and applications
Abstract: This paper proposes a novel adaptive control framework for nonlinear parametric-strict-feedback systems with a single unknown parameter entering nonlinearly. Instead of realizing practical tracking of reference signals which is already solved with existing methods, the proposed framework can achieve asymptotic tracking control. The design is divided into two steps. In the first step, an input-to-state stable (ISS) controller is designed, which renders the controlled system ISS with respect to the identification error and its time derivative. In the second step, a novel estimator is proposed. It is theoretically proved that global asymptotic stability is reached when the controlled system is connected to the estimator.

Keywords: Autonomous vehicles, Deep learning and machine learning, Adaptive systems
Abstract: GNSS/INS integrated navigation provides stable navigation performance by combining GNSS information with continuous inertial information from INS. Although low-cost MEMS IMUs are widely used, their inherent bias and noise cause error accumulation. To address this problem, virtual IMU (VIMU) and federated filter methods have been used. VIMU combines measurements from multiple IMUs into a single INS. Federated filters estimate states independently for each IMU and then combine the estimates. However, faults in some IMUs can distort the fused result and degrade navigation performance. This study proposes a multi-IMU fusion method based on interacting multiple model (IMM) within a federated filtering framework to maintain reliable state estimation performance under IMU fault conditions. To remove the effect of faulty IMUs, a deep deterministic policy gradient (DDPG) method is used to adjust the IMM mode probabilities according to the fault condition. The results show that the proposed method outperforms VIMU and conventional federated filtering under IMU fault conditions.

Keywords: Intelligent control, Nonlinear control and applications, Adaptive systems
Abstract: This article focuses on trajectory generation and adaptive tracking control for the flapping wing aerial vehicle with unknown dead zone. The unknown dead zone input is modeled as a time-varying nonlinear function, and appropriate design parameters are selected to effectively compensate for the impact of the unknown dead zone. On this basis, the neural networks are introduced to handle the impact of continuous unknown uncertainties caused by the inaccurate mass of the vehicle. Then, through Lyapunov stability analysis, it is proven that all closed-loop signals are bounded. Finally, the simulation results indicate that the proposed controller can effectively track the desired trajectory by selecting the appropriate control gain.

Keywords: Motion and vibration control, Nonlinear control and applications, Adaptive systems
Abstract: This paper investigates the adaptive predefined-time attitude tracking problem for spacecraft subject to input delay, inertia uncertainties, and external disturbances. To effectively compensate for the influence of input delay on control performance, an auxiliary compensation system with predefined-time convergence characteristics is constructed in this paper. Based on this mechanism, a non-singular adaptive predefined-time attitude tracking control scheme is established via the backstepping method. In particular, the singularity issue associated with fractional-power feedback is avoided through a hyperbolic tangent based construction, while an adaptive compensation term is designed to suppress the lumped uncertainties induced by inertial parameter variations and disturbances. Rigorous analysis shows that all signals in the closed-loop system remain bounded and the desired predefined-time convergence behavior is preserved. Comparative simulation studies further confirm the superior tracking performance and enhanced delay-tolerance capability of the proposed scheme.

Keywords: Energy Systems
Abstract: In this paper, we address the problem of benefit maximization from energy consumption of a curtailment-aware large prosumer by leveraging employees' electric vehicles. First, we introduce a model of the prosumer with electric vehicles, which reflects practical regulatory constraints on curtailments. We further define the prosumer's benefit, taking into account not only monetary benefit but also greenness-aware benefit through the decentralized energy systems investment. Then, we formulate and solve a maximization problem of prosumer benefit. A solution to this problem provides an optimal charging/discharging plan of electric vehicles that reduces both variable renewable energy curtailments and demand curtailments. The effectiveness of the proposed framework is illustrated using realistic data through a case study of a hospital acting as a large prosumer.

Keywords: Energy Systems, Industrial applications
Abstract: Capacity planning of electricity-hydrogen integrated energy system faces growing challenges from the behavior-driven uncertainty of electric vehicle charging. This paper proposes a learning-enhanced two-stage distributionally robust planning framework to address this challenge. First, electric vehicle charging behavior is characterized through grouped and locally adaptive kernel density estimation, capturing its multi-peak and heterogeneous patterns. Then, a machine-learning-based uncertainty set construction method is developed. By integrating density-based spatial clustering of applications with noise, Gaussian mixture modeling, and principal component analysis, the method extracts structural information from multi-source historical data while suppressing the influence of low-probability outliers. The resulting structured support subsets are embedded into a two-stage distributionally robust planning model that incorporates flexible electric vehicle charging. Case studies in an industrial park demonstrate that the proposed machine-learning-based support set reduces the annual total cost by 3.07% compared to a conventional support set, while flexible charging further improves operational economy. These results confirm that the combination of learning charging behavior and uncertainty structure enhances both the economic efficiency and the robustness of long-term capacity planning.

Keywords: Energy Systems, Industrial applications
Abstract: High renewable penetration makes transmission expansion planning increasingly sensitive to operating uncertainty. A plan optimized only for a nominal operating point may remain feasible, yet perform poorly in renewable accommodation when load levels and renewable outputs deviate from their forecasted values. This paper proposes a scenario-based robust transmission expansion planning model for renewable accommodation under load and renewable uncertainty. The model adopts a mixed-integer linear programming formulation with shared line-investment decisions and scenario-dependent DC dispatch variables. Renewable curtailment and load shedding are explicitly modeled, and the planning objective balances transmission investment and multi-scenario operational performance in a unified framework. Numerical studies show that the robust plan adds one extra candidate corridor compared with the deterministic plan, increasing investment cost from 25 to 31, while reducing average renewable curtailment under the same planning scenarios from 75.050 MW to 68.500 MW. In out-of-sample evaluation on 20 random scenarios, the robust plan further reduces average curtailment from 59.774 MW to 53.737 MW and worst-case curtailment from 149.985 MW to 139.805 MW, without introducing load shedding. These results indicate that a modest increase in transmission investment can improve renewable accommodation and enhance the adaptability of the expansion plan to operating uncertainty.

Keywords: Energy Systems, System identification and modelling, Communication
Abstract: The estimation precision of state of charge(SOC) and state of health(SOH) for lithium iron phosphate(LFP) batteries, based on equivalent circuit models, is easily compromised by voltage measurement errors arising from the flat open circuit voltage(OCV) curve. To mitigate the issue, a joint SOC and SOH estimation method using an error-compensated extended Kalman filter (EC-EKF) is presented. A first-order RC ECM is established, and the factors contributing to estimation errors are analyzed. Capacity and error terms are incorporated into the state-space equations for observation, and the EKF is employed for state estimation. The capacity estimated at the transition point between the voltage plateau and the steep OCV region is used as the initial capacity for the next cycle. The closed-loop property of EKF enables accurate SOC and voltage error estimation, with the estimated voltage error compensating SOC. Experimental results under various dynamic conditions validate the feasibility of the proposed method.

Keywords: Energy Systems, Deep learning and machine learning, Artificial intelligence
Abstract: This paper presents a risk-aware framework for optimal microgrid scheduling using distributional reinforcement learning and mixed-integer programming (MIP). The proposed method integrates distributional value estimation with conditional value-at-risk (CVaR)-based decision making to explicitly capture uncertainty and operational risk. Unlike conventional deep reinforcement learning (DRL) methods that optimize only the expected return, the proposed approach learns the distribution of cumulative returns and enables risk-sensitive dispatch decisions. Operational constraints including power balance, generator ramping limits, and energy storage system (ESS) state-of-charge (SOC) bounds are handled through an MIP-based scheduling framework. The method is evaluated on a grid-connected microgrid with distributed generators (DGs), photovoltaic (PV) generation, ESSs, and stochastic electricity demand. Numerical results demonstrate that the proposed framework improves robustness under uncertainty and provides an effective solution for reliable microgrid scheduling.

Keywords: Energy Systems, Industrial applications, Autonomous vehicles
Abstract: This paper presents a battery degradation–aware route planning framework for electric vehicles (EVs) through a practical case study of the Bandung–Cirebon corridor in West Java, Indonesia, a route characterized by complex topography, steep elevation changes, and heterogeneous road surface conditions. An integrated routing cost function is developed by combining energy consumption, derived from a physics-based vehicle dynamics model, with vibration-induced battery stress metrics. To capture the influence of road geometry and surface conditions, the cost function is applied to a 4D road network representation that includes lateral and longitudinal geometry, elevation profiles, and road-induced vibration characteristics. Route optimization is performed using a Dijkstra-based routing framework. The results demonstrate that the proposed approach successfully redirects EVs toward routes that reduce cumulative battery stress while maintaining reasonable travel efficiency. The findings highlight the importance of incorporating topographic and vibration-related factors into EV routing strategies to mitigate accelerated battery degradation in challenging road environments.

Keywords: Energy Systems, System identification and modelling, Industrial applications
Abstract: With the continuous expansion of the scale of power grids and the increasing complexity of operating conditions, objectively assessing operational risks in power systems to support dispatch decision-making has become a critical issue. To address the limitations of conventional assessment methods that rely heavily on expert experience and lack quantitative characterization of risk consequences, this paper proposes a power system operational risk assessment method based on topological scenario evolution and bottleneck constraints. The proposed approach constructs system-wide N-1 contingency risk scenarios using equipment operating states and network topology information, and employs depth-first search (DFS) to automatically identify the cascading impact range of faults, thereby generating a systematic risk scenario set. Furthermore, a risk quantification model is developed by introducing a worst-performing-indicator-driven bottleneck constraint mechanism, which enhances the sensitivity of the assessment results to severe risk consequence scenarios. Finally, grouped scenario statistics combined with the Best–Worst Method (BWM) are applied to achieve quantitative system-level situational assessment. Case studies on a practical power system demonstrate that the proposed method can effectively identify structurally high-risk operating scenarios,providing strong support for grid operational situational analysis and security early warning.

Keywords: System identification and modelling, Energy Systems
Abstract: This paper presents a data-driven framework for nonlinear system identification using a dual Kalman filter that simultaneously estimates system states and model parameters. The proposed approach incorporates two independent forgetting factors, allowing adaptive tracking of both state dynamics and parameter variations while improving robustness to modeling uncertainties and time-varying behaviors. The method is applied to the discovery of nonlinear models for permanent magnet synchronous motors (PMSMs), a class of systems widely used in high-performance electric drives. By leveraging measured data and recursive estimation, the framework identifies nonlinear relationships without requiring a fully predefined physical model. Simulation results demonstrate promising performance in accurately capturing the nonlinear dynamics of the PMSM, highlighting the effectiveness of the dual Kalman filter for data-driven nonlinear modeling and its potential for applications in adaptive control and intelligent motor diagnostics.

Keywords: Control theories, Cyber-physical systems and security
Abstract: In this paper, we consider the privacy preservation problem in a cooperation of LQG control system. In this system, the user sends the innovation to the server, which then calculates the optimal control input. However, transmitting the innovation directly can easily lead to privacy leakage. Moreover, the server equips a covariance detector to check the statistics of the received innovations sequence, which makes most existing privacy preservation methods that usually alter the covariance of the original innovations fail to serve. To address the above issues, this paper proposes a novel local privacy preservation scheme. Firstly, by performing a properly designed affine transformation on the innovation values, the generated privacy signals are able to maintain identical covariances, thereby effectively avoiding the detection of the server. Secondly, a recovery mechanism is designed to ensure that the user can obtain accurate posterior estimates and optimal control inputs, thereby achieving privacy preservation without sacrificing control performance. Thirdly, this paper proposes a novel privacy metric, analyzes the privacy preservation quality of closed-loop systems, and proposes an optimization problem to maximize the protection on privacy. Finally, through the theoretical proof and the simulation result, the analysis is conducted on when the system can achieve the maximum privacy preservation performance.

Keywords: Cyber-physical systems and security
Abstract: This paper investigates a privacy-aware transmission control strategy for a hybrid automatic repeat request (HARQ)-based remote state estimation system. In this framework, the reliability of packet reception is improved due to HARQ soft combining as the number of consecutive retransmissions increases. Upon a transmission failure, a decision must be made regarding whether to retransmit the previously failed local estimate or to replace it by transmitting a newly generated estimate. A fundamental trade-off between transmission reliability and data freshness is induced by this decision. Furthermore, as transmissions are repeated, information about the system state can be intercepted and gradually accumulated by an external eavesdropper, thereby increasing privacy leakage risks. Consequently, a complex trade-off among data freshness, transmission reliability, and privacy preservation is introduced. By jointly evaluating the privacy leakage and the state estimation error, the transmission decision problem is formulated as a Markov decision process (MDP). To solve this complex problem, a model-free reinforcement learning (RL) approach based on Q-learning is proposed to derive the optimal transmission policy. Finally, numerical results are provided to demonstrate that data freshness and privacy are effectively balanced by the proposed Q-learning algorithm, and that traditional baseline policies are outperformed even in severe eavesdropping environments.

Keywords: Deep learning and machine learning, Autonomous vehicles
Abstract: The paper proposes a topology-guided multimodal spatiotemporal fusion framework for reinforcement learning(RL)-based autonomous aerial vehicle(AAV) navigation in unknown and complex environments, with the aim of reducing navigation failures caused by local optima. Existing navigation policies rely primarily on instantaneous local observations and often struggle to maintain stable and effective motion when goal direction is blocked by terrain or obstacles. To address the limitation, connected traversable regions are extracted from depth images to construct candidate topological nodes, and temporally consistent topology-guided direction is generated through historical node association. The guidance signal is fused with depth images, AAV states, and goal representations within the multimodal recurrent policy, while directional alignment loss is introduced to strengthen structural awareness and improve long-horizon decision-making under partial observability. Experimental results demonstrate improved navigation robustness and stronger trap-escaping capability in cluttered environments.

Keywords: Autonomous vehicles, Industrial applications
Abstract: With the rapid development and application of autonomous driving and artificial intelligence technologies, unmanned vehicles, as embodied intelligence agents, can construct different QoE models for vehicle users (VUs). These models can support VUs to request video streaming services based on their diverse preferences for factors such as resolution, frame rate, and rebuffering duration. However, most existing methods model the preferences of VUs using fixed weights, which makes it difficult to comprehensively account for factors such as high-speed vehicle mobility, road scene switching, and wireless link fluctuations, thus failing to characterize the evolution of the preferences in dynamic environments. To address this issue, this paper proposes an embodied intelligence-based preference perception scheme for video streaming services. Specifically, this paper first designs a static QoE modeling method based on experimental data, establishing a nonlinear mapping between various influencing factors and QoE to effectively characterize the baseline preferences of VUs in static scenarios. Then, we propose a dynamic preference adjustment method integrating environmental awareness and sensitivity analysis to enable the QoE preferences of VUs to evolve with environmental dynamics. Experimental results show that the proposed scheme can effectively reflect the dynamic evolution of the preferences of VUs in dynamic environments.

Keywords: Deep learning and machine learning, Artificial intelligence, Industrial applications
Abstract: For fingerprint-based indoor localization, variations in spatial structures and signal characteristics in different scenarios cause significant distribution shifts in fingerprint data, limiting the generalization of models trained in a single environment. Meanwhile, fingerprint data are distributed across devices or institutions and cannot be centrally aggregated due to privacy constraints. To address these issues, this paper proposes a federated meta-learning framework with analogical modeling (FML-AL), which integrates federated learning and meta-learning while modeling relative spatial relationships among samples. The localization model consists of a semantic feature encoder, a relative feature extractor, and a dual-branch location predictor. By jointly optimizing absolute and relative localization tasks, the proposed framework can improve generalization across heterogeneous scenarios and enables fast adaptation to new environments. Experimental results demonstrate that FML-AL outperforms conventional methods in both localization accuracy and scenario adaption performance.

Keywords: Cyber-physical systems and security
Abstract: This paper investigates the privacy-preserving average consensus problem of single-integrator systems, in which the initial values need to be protected through information interaction. Unlike the traditional state decomposition approaches, our method employs a state transition approach that stores the initial values in the internal substates, which serve as inputs to the external substates. Thus, the information interaction among external substates enables the agents to achieve average consensus and the internal states preserve the initial values. We propose a noise-cloaked dynamic average consensus algorithm which not only enhances privacy protection but also ensures that the states of agents converge to the average of the initial values over detail-balanced digraphs. Numerical simulations validate the approach and illustrate the tunable trade-off between privacy and convergence.

Keywords: Energy Systems
Abstract: This paper addresses the distributed operation strategy optimization problem for multiple battery energy storage systems (BESSs) in a smart distribution network under measurement noise uncertainties. To achieve supply–demand balance with user‑side multi‑battery BESSs, a robust estimation‑based optimal operation scheme is proposed. First, a pre‑designed distributed robust state estimation method is employed to obtain robust estimates of load power, mitigating the impact of complex noises. Then, a distributed operation strategy is derived by minimizing an objective function that simultaneously considers both the supply–demand mismatch and the charging efficiency loss. Finally, simulations on a standard bus system validate the scalability and feasibility of the proposed distributed control algorithm, demonstrating its effectiveness in balancing supply–demand and reducing charging losses under noisy conditions.

Keywords: Cyber-physical systems and security, Control theories, Control devices, sensors and actuators
Abstract: This paper investigates the security of remote state estimation systems subjected to physical False Data Injection (FDI) attacks under resource constraints. Unlike the unconstrained scenario, the attacker can only inject malicious data intermittently due to limited resources, which is characterized by a Bernoulli binary variable. We have determined the necessary and sufficient conditions for the attacker to maintain stealthiness against residual detector while causing the bias in the physical process state to expand infinitely. It is worth noting that, compared to the attack scenario with sufficient resources, the necessary and sufficient conditions in the scenario considered in this paper impose stricter constraints on the attack strategy. Numerical simulations are ultimately provided to validate the theoretical results that have been developed.

Keywords: Communication, Nonlinear control and applications, Control theories
Abstract: This paper addresses the semantic-triggered tracking control problem for nonlinear systems subject to output constraints. A novel predictive model-based semantic-triggered mechanism is constructed, which leverages contextual data to decide transmission events. Subsequently, semantic extraction techniques are used to distill semantic triplets from the high‑dimensional raw data for transmission, thereby conserving communication resources by reducing both the update frequency and the data volume. To handle semantic errors and output constraints, a recursive backstepping controller is developed within a Barrier Lyapunov function (BLF). This design ensures that all closed-loop signals are semi-globally uniformly ultimately bounded while strictly enforcing the output constraints.

Keywords: Control theories, Cyber-physical systems and security, Control devices, sensors and actuators
Abstract: This paper investigates the event-triggered output-feedback model predictive load frequency control problem in smart grids. To manage the transmission of measurement data packets, a novel dynamic event-triggered mechanism incorporating an adjustment variable and an internal dynamic variable is proposed. The dynamic output-feedback model predictive control problem is formulated as a ``min-max'' optimization problem. By constructing a Lyapunov-like function with an internal dynamic variable, the original optimization problem is reformulated as an auxiliary one constrained by a set of matrix inequalities. A systematic design method for the output-feedback controller is presented. Theoretical analysis demonstrates the feasibility of the algorithm and the stability of the system. Finally, the effectiveness of the proposed strategy is validated through a case study.

Keywords: Cyber-physical systems and security, Control theories, Deep learning and machine learning
Abstract: We study resilient learning for heterogeneous multiagent multiarmed bandits (MA2Bs) in the presence of Byzantine agents. In heterogeneous MA2Bs, inter-agent communication is essential for learning the globally optimal arm, but it also makes the learning process vulnerable to Byzantine attacks. Moreover, unlike homogeneous settings, directly trimming extreme information may discard useful observations from normal agents, and simply isolating detected Byzantine agents cannot remove the residual impact of manipulated information injected before detection. To address this issue, we propose a Byzantine-resilient distributed UCB algorithm that mitigates the historical influence of Byzantine information on the global reward mean estimates. Theoretical analysis shows that the proposed algorithm enables each normal agent to achieve the same asymptotic logarithmic regret bound as in the attack-free case after correct detection of Byzantine neighbors. Simulation results further demonstrate that simple post-detection isolation may still suffer severe performance degradation, whereas the proposed algorithm effectively recovers the learning performance.

Keywords: Cyber-physical systems and security, Control theories
Abstract: This work explores the robust synchronization of coupled Boolean networks (BNs) under stuck-at fault conditions. Since these faults fix node states and invalidate original synchronization criteria, we propose a novel fault-preserving subset to describe the resulting invariant state evolution. Leveraging this concept, we establish necessary and sufficient conditions to determine the maintenance of synchronization, notably avoiding the reconstruction of the faulty model. Numerical simulations validate the theoretical findings.

Keywords: Cyber-physical systems and security, Control theories, Control devices, sensors and actuators
Abstract: This paper investigates a reduced-order observability analysis framework for critical sensor identification in large-scale cyber-physical systems (CPSs). Due to the high dimensionality of such systems, direct computation of observability-related metrics is often computationally expensive. To address this issue, a Petrov--Galerkin (PG) projection-based model reduction method is introduced to construct a lower-dimensional model while preserving the main observability-related characteristics of the original system. Based on the reduced-order model, the degree of observability is quantitatively evaluated by the trace metric of the observability Gramian. In addition, the approximation error between the reduced-order and full-order observability Gramians is analyzed, and a computable upper bound is derived. The proposed framework provides an efficient way to assess the sensing effectiveness of different sensor channels and identify the most important sensors in large-scale systems. Numerical results demonstrate the effectiveness of the proposed method.

Keywords: Cyber-physical systems and security, Intelligent control, Artificial intelligence
Abstract: As the information layer and the physical layer become increasingly integrated, active distribution networks tightly couple communication, sensing, and control information systems,which has given rise to Cyber-Physical Power Systems (CPPS). Furthermore, the high proportion of renewable energy integration has further increased the complexity of system operations. This has led to a growing threat of attacks on the information layer of power systems.Traditional defense methods, which rely on fixed rules and static thresholds, struggle to adapt to dynamically evolving attack scenarios.This paper proposes a rule and data driven self-evolving defence method (RDD-SGDM). The method integrates the interpretability of rule-based approaches with the adaptability of data-driven methods. It establishes a closed-loop mechanism of rule screening, data learning, and feedback updating. This enables continuous iterative optimisation of defence strategies while balancing security, stability and economic objectives.Experimental results show that RDD-SGDM significantly enhances system adaptability to information-layer attacks. Both defence efficacy and operational robustness are effectively improved. This work provides a promising solution for active defence in next generation cyber-physical power systems.

Keywords: Intelligent control, Deep learning and machine learning, Industrial applications
Abstract: Unmanned aerial vehicle (UAV)-based photovoltaic (PV) inspection is significantly bottlenecked by the conventional "one-task-one-model" paradigm, where cascading task-specific models causes high computational overhead and severe switching latency on edge devices. To challenge this, we propose a unified ultra-lightweight detection framework based on YOLOv8n-P2, advocating a "single-model full parameter sharing" paradigm. By leveraging a high-resolution P2 layer and a state-driven progressive workflow, a single network extracts a unified feature basis to solve multi-scenario tasks across disparate modalities: long-distance aerial localization, close-range visible inspection, and electroluminescence (EL) diagnosis. This design fundamentally eliminates the latency and memory fragmentation caused by hot-swapping heterogeneous models. Experiments demonstrate that the model, with only 3.35M parameters, achieves over 0.9 accuracy in macro-localization and EL detection tasks. Deployed with a custom digital twin interface, this framework enables real-time multi-modal inference, proving the viability of our zero-switching paradigm for closed-loop UAV edge inspection.

Keywords: Cyber-physical systems and security
Abstract: This paper proposes an encoding-decoding interval observer-based fault detection scheme for multi-agent systems subject to limited communication bandwidth. Specifically, an encoding-decoding mechanism is employed, with a novel scaling function designed to mitigate unbounded data growth caused by traditional methods under faulty conditions. Furthermore, the inevitable encoding errors are explicitly modeled and incorporated into the observer design framework. By satisfying mixed l_1/H_infty performance indices, the proposed scheme theoretically guarantees the natural threshold property, thereby ensuring reliable fault detection in communication-constrained environments. Finally, a simulation example is provided to verify the effectiveness of the proposed scheme.

Keywords: Autonomous vehicles, Control theories, Cyber-physical systems and security
Abstract: Platooning control of connected automated vehicles (CAVs) has emerged as a promising solution to alleviate traffic congestion, reduce fuel consumption, and improve safety for modern transportation systems. This paper addresses communication constraints induced by denial-of-service (DoS) attacks and limited bandwidth, as well as a performance constraint arising from actuator saturation. Under these constraints, a dynamic event-triggered mechanism (DETM) is adopted to reduce the communication burden, and a distributed robust model predictive control (MPC) approach is proposed to guarantee formal closed-loop system stability and H_infty performance. A constrained optimization problem under the distributed robust MPC framework is proposed to obtain feasible controller gains, which fully considers the characteristics of DoS attacks and actuator saturation. Finally, several numerical simulation results are provided to demonstrate the effectiveness of the derived theoretical results.

Keywords: Autonomous vehicles, Control theories, Intelligent control
Abstract: This paper addresses the problem of event-triggered adaptive platoon control for autonomous vehicles (AVs) in the presence of model uncertainties, input delays, and actuator saturation. To cope with unknown nonlinear dynamics and external disturbances, an observer-based adaptive framework enhanced with neural network approximation is developed, enabling accurate estimation and compensation of unmodeled effects. Meanwhile, a delay compensation scheme is incorporated to counteract the influence of input delays, thereby improving tracking performance and system robustness. In contrast to conventional platoon control designs, actuator saturation constraints are explicitly considered, and their impact on closed-loop stability is rigorously analyzed, leading to a control strategy that is more consistent with practical vehicle limitations. Furthermore, a relative-threshold event-triggered mechanism is proposed to regulate the update of control inputs for each AV. The triggering condition adapts to the magnitude of the control effort and effectively reduces unnecessary updates while preserving control accuracy. It is shown that the resulting closed-loop system guarantees stable leader–follower tracking and excludes Zeno behavior. The proposed approach also accommodates coupled longitudinal and lateral vehicle dynamics, allowing coordinated platoon motion in multi-lane scenarios. Numerical simulations are provided to demonstrate the efficacy of the proposed platoon control approach.

Keywords: Adaptive systems, Nonlinear control and applications
Abstract: This paper investigates a consensus tracking problem for high-order multi-agent systems (MASs). To eliminate the need for full-state information of the leader, a fixed-time reference system is designed for each agent by using only the reference signal and its first-order derivative exchanged over communication networks. Based on the reference system design, a command-filtered backstepping controller is proposed to achieve consensus tracking while avoiding the complexity explosion and singularity issues existed in recursive procedures. It is shown that stability is guaranteed for the MASs and all signals remain bounded. Numerical simulation results are provided to verify the effectiveness of the proposed control scheme.