Keywords: Optimization, Uncertain systems, Machine learning
Abstract: In this seminar, we start with a broad class of anomaly detection for large-scale nonlinear dynamical systems. Noting a connection between the diagnosis filter and the so-called behavioral sets of dynamical systems, we leverage tools from the traditional model-based approaches and modern data-driven analytics to address the inherent complexity of the problem. We then shift our attention to the performance guarantees of the proposed solution. In this part, we study this topic in a general context of data-driven decision-making with a particular focus on the distributionally robust optimization framework. We will discuss the role of convexity from the different viewpoints of computational, statistical, and real-time implementation.

Keywords: Automotive, Autonomous systems, Transportation systems
Abstract: This work investigated the effect of a Nu Vinci type Continuous Variable Transmission (CVT) based Energy Efficient Cruise Control (EECC) on an Electric Vehicle's (EV's) energy consumption. Unlike petrol and diesel vehicles, almost all EVs have a fixed gear ratio. Although it reduces the capital cost, it may not offer optimal energy efficiency and may increase the recharge costs. While accelerating and cruising, it may cause an EV to spend more energy than needed. While braking, it may cause underutilisation of the regenerative braking potential. In this work, an EECC was designed to operate the EV's power train close to its peak efficiency region by controlling the CVT ratio and Electric Machine (EM) torque. The EECC exploits data from the lead vehicle and employs constrained linear Model Predictive Control (MPC) method with a novel problem formulation that reduces the computational complexity. The proposed EECC was tested in a MATLAB simulation environment for different drive cycles. The results show that compared to a baseline EV with a fixed gear ratio and an Adaptive Cruise Control, the proposed system can reduce an EV's energy consumption in urban drive cycles by 16.6%.

Keywords: Automotive, Predictive control for linear systems, Robust control
Abstract: The autonomous vehicle following problem has been extensively studied for at least two decades with the rapid development of intelligent transport systems. In this context, this paper proposes a robust model predictive control (RMPC) method that aims to find the energy-efficient following velocity of an ego battery electric vehicle and to guarantee a safe rear-end distance in the presence of disturbances and modelling errors. The optimisation problem is formulated in the space domain so that the overall problem can be convexified in the form of a semi-definite program, which ensures a rapid solving speed and a unique solution. Simulations are carried out to provide numerical comparisons with a nominal model predictive control (MPC) scheme. It is shown that the RMPC guarantees robust constraint satisfaction for the closed-loop system whereas constraints may be violated when the nominal MPC is in use. Moreover, the impact of the prediction horizon length on optimality is investigated, showing that a finely tuned horizon could produce significant energy savings.

Keywords: Automotive, Transportation systems, Sliding mode control
Abstract: Road vehicle platoons improve fuel economy by limiting the aerodynamic drag. Time-headway along with operating speed dictates the intervehicular spacing, and hence the aerodynamic drag reduction. This paper proposes an adaptive scheme to adjust the magnitude of time-headway to reduce energy consumption according to operating speed. An ‘energy consumption - time-headway - speed map’ is generated using a complete vehicle dynamics-based platoon framework integrated with a sliding mode controller. This map could be used to adaptively adjust time-headway for an expected energy consumption magnitude and to predict the energy consumed for the completion of a route. Constraints are enforced limiting the time-headway values to meet deceleration demand. The proposed approach was found to result in an energy consumption reduction up to 35% compared to conventional approaches where the time-headway values are kept constant.

Keywords: Automotive, Identification for control, Mechatronics
Abstract: Modern agricultural vehicles are complex machines that require many automatic controls. Combine harvesters are particularly challenging from a control perspective. A critical control problem in combines is chassis leveling. Keeping the chassis level with respect to the gravity enhances both the crop processing efficiency and the vehicle safety. This paper presents a leveling control algorithm based on switching hydraulic actuators. A gray-box system identification is first performed, and the regulator is designed as a two-layer cascade scheme. Closed-loop experimental tests on the real vehicle prove the effectiveness of the designed controller, showing a reduction of 40% in terms of root mean square error of the roll angle, with respect to an uncontrolled case.

Keywords: Automotive, Modeling, Robust control
Abstract: Model Free Control (MFC) with ultra-local model is a novel design method to provide high-performance control for systems with nonlinear dynamics. The basic concept of this approach is to compute an ultra-local model, which is valid for a short period of time. Although the concept has already been successfully applied for various vehicle control problems, there are some open issues regarding the robustness and the stability of the closed-loop controlled system. This paper presents a method for designing a robust control based on H8 synthesis, with which a solution to these issues can be found. In the proposed MFC strategy the ultra-local model-based control is in the inner loop of the control structure. Moreover, a robust H8 control for the outer loop is designed, in whose control-oriented model the control of the inner loop is incorporated. The control strategy is implemented in a high-fidelity vehicle dynamic environment and its efficiency and operation are demonstrated through a complex simulation example.

Keywords: Automotive, Optimal control, Switched systems
Abstract: In this paper, a complete eco–driving strategy for heavy–duty trucks (HDT) based on a finite number of driving modes with corresponding gear shifting is developed to cope with different route events and with road slope data. The problem is formulated as an optimal control problem with respect to fuel consumption and trip duration, and solved using a Pontryagin minimum principle (PMP) algorithm for a path search problem, such that computations can be carried out online, in real–time. The developed eco—driving assistance system (EDAS) provides a velocity profile and a sequence of driving modes (and gears) recommendation to the driver, without actively controlling the HDT (human in the loop) and, in practice, allows contextual feedback incorporation from the driver for safety. Simulation results show that the developed methodology is able to provide a velocity profile for a complete route based on known road events and slope information while satisfying all truck operational constraints.

Keywords: Automotive, Predictive control for nonlinear systems, Optimal control
Abstract: Vehicle thermal management system is a key facilitator of driving range increase of a battery electric vehicle in extreme weather conditions. This control system should optimally coordinate control actions of heating, ventilation, and air-conditioning (HVAC) system to maintain high thermal comfort while reducing energy consumption. To this end, this paper presents a nonlinear model predictive control (NMPC) system for electric vehicle cabin cooling system. The proposed NMPC strategy optimizes trajectories of cabin inlet air temperature and air mass flow, while accounting for cabin thermal dynamics, available disturbance preview, and low-level-controlled HVAC system’s dynamics and operating range constraints. The NMPC cost function includes simultaneous minimization of Predicted Mean Vote (PMV) thermal comfort index and either maximization of HVAC coefficient of performance or minimization of HVAC power consumption. The proposed NMPC system is verified through simulation for cabin cool-down scenario and compared with a hierarchical control strategy based on a superimposed cabin air temperature feedback controller and an inner control allocation algorithm. The simulation results indicate that the NMPC system outperforms the hierarchical control strategy in terms of reduced energy consumption for the same comfort or improved thermal comfort for the same energy consumption.

Keywords: Distributed control, Predictive control for linear systems
Abstract: Coalitional control is a type of distributed control characterized by the dynamic adjustment of the overall controller structure so that only strongly coupled agents interact with each other. In particular, local controllers merge into cooperative coalitions (or clusters) only when it improves global performance, thus reducing the overall cooperation burden. This paper proposes a novel coalitional model predictive control (MPC) approach in which coupled variables are decomposed into a public part, which is optimized by the neighboring agents, and a private one, which is locally controlled by the agent that owns it. The bounds on these variables are negotiated in a distributed manner, and a threshold is established to trigger different interaction modes, including the classical decentralized and distributed approaches, and flexible modes of cooperation. Finally, to illustrate the benefits of this control scheme, results on a simulated eight input-coupled tanks plant are provided.

Keywords: Predictive control for nonlinear systems, Autonomous systems, Machine learning
Abstract: Safe, collision-free movement of autonomous sys- tems such as robots, mobile platforms, or drones in cluttered environments is challenging. Often the exact positions and di- mensions of other systems and objects are uncertain. However, data from successful, previous trajectories might be available. We design a learning-supported model predictive controller for autonomous systems to navigate through a “safe” path- corridor learned from prior collision-free movement trajectories using Gaussian process regression. The posterior mean and variance of the Gaussian process define a corridor that allows for safe transition through the cluttered environment. A model predictive controller is used to find the optimal path inside the learned corridor and steers the system to follow it. It guarantees satisfaction of constraints on the system as well as on the reference path which is subject to the learned corridor limitations. Simulation studies for an autonomous mobile robot that navigates through an environment with obstacles demon- strate the approach’s benefits. It is shown that the controller’s flexibility to move freely in the safe path corridor increases the performance when compared to using a predefined fixed path.

Keywords: Observers for linear systems, Predictive control for linear systems
Abstract: This paper introduces a data-based moving horizon estimation (MHE) scheme for linear time-invariant discrete-time systems. The scheme solely relies on collected data without employing any system identification step. Robust global exponential stability of the data-based MHE is proven under standard assumptions for the case where the online output measurements are corrupted by some non-vanishing measurement noise. A simulation example illustrates the behavior of the data-based MHE scheme.

Keywords: Optimal control, Energy systems, Modeling
Abstract: This article presents the economic optimization of a refrigeration energy supply system based on air-cooled water chillers, to which an energy storage system (TES) has been added. This is to satisfy the cold demand of a large facility. The modelling of each of the elements has been integrated into Simulink and Simscape in order to be able to simulate the complete plant. It has been assumed that the system is controlled at low level, in such a way that is capable of receiving flow and temperature references from an optimizer placed hierarchically on a higher level. Finally, a non-convex mixed-integer optimization problem has been proposed with continuous and binary variables associated to the operation of parts of the facility, so that the economic cost of cold generation is optimized. Simulation results that support the good performance of the optimizer are presented.

Keywords: Predictive control for nonlinear systems, Fuzzy systems
Abstract: This paper presents the design of a Model Predictive Control (MPC) strategy for Takagi & Sugeno (TS) systems that is based on a control law with optimised prediction dynamics, first proposed in a context of Robust MPC for systems with multiplicative uncertainty. Based on the similarities between this kind of systems and state-space TS systems, this predicted control law is adapted to fuzzy models to exploit the known information of the normalised degrees of activation. It is described how to design the parameters of the controller and how to apply it closed-loop fashion. It is shown that the proposed controller is guaranteed to be recursively feasible and asymptotically stabilises the controlled systems. A simulation example shows the attributes and benefits of the proposed controller.

Keywords: Electrical machine control, Computational methods, Predictive control for nonlinear systems
Abstract: In this paper, a numerical solution approach for the field oriented economic model predictive control (FOEMPC) of permanent magnet synchronous machines (PMSM) is proposed. The FO-EMPC is based on the dq-model of the PMSM with spherical voltage constraint. A terminal set and a terminal penalty are used to mitigate stability and convergence problems with a short prediction horizon. The resulting optimal control problem (OCP) is formulated as a nonlinear program (NLP) and is solved with a sequential quadratic programming (SQP) scheme. Considering the structure of the system and the objective function, the quadratic problem (QP) is solved efficiently with an adapted condensing method and by using the Schur complement of the Karush Kuhn Tucker (KKT) matrix. This leads to small achievable computation times and allows the use of the model predictive control for electric drives on a standard automotive electronic control unit (ECU). An experiment on a motor test bench demonstrates the efficiency of the proposed numerical framework and shows the superior control performance of the FO-EMPC, which is significantly better than state-of-the-art approaches.

Keywords: Process control, Energy systems, Predictive control for nonlinear systems
Abstract: Line focus concentrating solar collectors control operates by manipulating the flow for tracking the desired outlet temperature under normal conditions. This solar collector can also use the solar tracking device to manipulate the mirror's angle for defocusing under abnormal events. Defocusing mirrors is the last control measure to avoid overheating because defocusing means wasting energy. This work proposes changing the defocus use as the last control resort for safety, to a standard manipulated variable combined with the flow for outlet temperature tracking. The proposal uses a multi-variable non-linear MPC technique with a modified objective function and simulates three scenarios on a Fresnel solar collector. Besides, this paper indicates that the defocus action is necessary for regular operation; thus, it is also essential to consider the defocus as a manipulated variable for controller design. The proposed controller tracks reference and rejects disturbances while having defocus minimization, overheating prevention, and thermal power reference tracking features.

Keywords: Decentralized control, Distributed cooperative control over networks, Distributed estimation over sensor nets
Abstract: The behaviour of most modern multi-agent networked systems, used for distributed learning and control tasks, is describable by a set of interacting algebraic Riccati equations. However, due to the complexity of their behaviour, these interacting Riccati equations have, to this point, not been subject to rigorous scrutiny. To this end, a general class of algebraic Riccati equations is considered, their behaviour is analysed, and conditions for convergence to a unique set of stabilising solutions is established. The class of algebraic Riccati equations considered in this work is selected so that obtained results would be generalisable for a wide range of statistical learning and control purposes. Finally, application of the obtained results in distributed Kalman filtering and decentralised linear control is demonstrated.

Keywords: Distributed cooperative control over networks, Communication networks, Network analysis and control
Abstract: Distributed event-triggered control for discrete-time stochastic multi-agent systems with non-uniform Bernoulli packet loss is investigated in this paper. An event-triggered strategy is proposed to reduce the load on the communication network, which can be an issue when the bandwidth is small. The proposed trigger condition requires only locally available information, thus enabling a distributed control scheme. To enhance the performance, a co-design of event-triggered strategy and controller gains is proposed. Agent dynamics (single integrators) are modeled in discrete time. The network is represented as a Markovian Jump Linear System, and sufficient conditions for mean square consensus are given in the form of a linear matrix inequality. Since this LMI condition can be a potential of huge size (when the network is large), we show how the synthesis condition can be turned into a robust synthesis problem with a complexity that is independent of the network size and corresponds to a synthesis condition for a single agent. A numerical example illustrates the approach and shows how the trigger level can be chosen to trade performance against data transmission rate.

Keywords: Distributed cooperative control over networks, Control over networks, Network analysis and control
Abstract: In this paper, we consider a physically-interconnected multi-agent system consisting of actuated and unactuated agents. The coordination of such a system seeks to ensure that all agents reach consensus through appropriate control of only the actuated agents. We propose a three-stage distributed control structure which achieves coordination by supplying the actuated agents with a common setpoint based on measurement information from both actuated and unactuated agents. This control structure comprises local PI controllers and dynamic distributed averaging with nearest neighbour communication. The stability of the closed loop is analysed by applying a generalized Kron reduction to construct a Lur'e problem. By applying the circle criterion, an LMI is proposed which verifies the absolute stability of the system. Finally, the closed-loop performance and robustness are illustrated via simulations.

Keywords: Distributed control, Energy systems, Emerging control applications
Abstract: The energy network is experiencing a decentralizing process due to the recent inclusion of distributed energy resources (DERs), such as photovoltaics (PV), electric vehicles (EV), or batteries. This paradigmatic shift involves new challenges that require enhancing the electricity system's flexibility while preserving its integrity and stability. One way of dealing with them is by unbundling the electrical network into smaller, more manageable units, known as microgrids (MG). Nonetheless, the usage of distributed architectures leads to more exchanged data and control information between end agents, involving security or privacy issues. In this context, blockchain technology emerges as a feasible solution that promises transparent, tamper-proof, and safe systems in a decentralized ecosystem. This paper proposes a distributed energy management platform that takes full advantage of the blockchain technology to enable safe peer to peer (P2P) transactions. A Distributed model predictive control (DMPC) scheme is adopted, and its performance is compared with the centralized approach.

Keywords: Distributed control, Optimization, Control over networks
Abstract: In this paper, we consider the problem of containment control of multi-agent systems with multiple stationary leaders, interacting over a directed network. While, containment control refers to just ensuring that the follower agents reach the convex hull of the leaders’ states, we focus on the problem where the followers achieve a consensus to the average values of the leaders’ states. We propose an algorithm that can be implemented in a distributed manner to achieve the above consensus among followers. Next we optimize the convergence rate of the followers to the average consensus by proper choice of weights for the interaction graph. This optimization is also performed in a distributed manner using Alternating Direction Method of Multipliers (ADMM). Finally, we complement our results by illustrating them with numerical examples.

Keywords: Distributed cooperative control over networks, Agents and autonomous systems, Nonlinear system theory
Abstract: This paper proposes a passivity approach in port- Hamiltonian (pH) form for multi-agent formation control and velocity tracking. The control law consists of two parts, where the internal feedback is to track the velocity and the external feedback is to achieve formation stabilization. Since the dynamics of the controller are associated with the edges, the stability analysis is related to the kernel of the incidence matrix B of the underlying graph. For displacement-based formations, the approach is applicable not only to acyclic graphs but also to cyclic graphs, in which case the columns of B are not linearly independent. For rigid formations, the passivity approach for distance and bearing formation is proposed. In addition, the relationship between infinitesimal rigidity and convergence of the desired formation is established. The proposed approach is verified by numerical simulations.

Keywords: Decentralized control, Computer networks, Network analysis and control
Abstract: We discuss control of bittide distributed systems, which are designed to provide logical synchronization between networked machines by observing data flow rates between adjacent systems at the physical network layer and controlling local reference clock frequencies. We analyze the performance of approximate proportional-integral control of the synchronization mechanism and develop a simple continuous-time model to show the resulting dynamics are stable for any positive choice of gains. We then construct explicit formulae to show that closed-loop performance measured using the L2 norm is a product of two terms, one depending only on resistance distances in the graph, and the other depending only on controller gains.

Keywords: Markov processes, Stochastic control, Machine learning
Abstract: This paper discusses the functional stability of closed-loop Markov Chains under optimal policies resulting from a discounted optimality criterion, forming Markov Decision Processes (MDPs). We investigate the stability of MDPs in the sense of probability measures (densities) underlying the state distributions and extend the dissipativity theory of Economic Model Predictive Control in order to characterize the MDP stability. This theory requires a so-called storage function satisfying a dissipativity inequality. In the probability measures space and for the discounted setting, we introduce new dissipativity conditions ensuring the MDP stability. We then use finite-horizon optimal control problems in order to generate valid storage functionals. In practice, we propose to use Q-learning to compute the storage functionals.

Keywords: Neural networks, Stability of nonlinear systems
Abstract: The formalism of Integral Quadratic Constraints (IQCs) is well-established in robust control. It has recently been used for systems with a neural network as one of its components, however, using only a small subset of the established techniques for obtaining IQC relations. We provide a larger set of IQCs relevant for the nonlinearities commonly used in neural networks, introduce new constraints for the rectified linear unit and the leaky rectified linear unit, and draw on the established literature to build a library of IQCs to use in connection with neural networks. Finally, our examples show how improved guarantees can be obtained with a larger library of IQCs.

Keywords: V&V of control algorithms, Safety critical systems, Statistical learning
Abstract: Hamilton-Jacobi reachability methods for safety-critical control have been well studied, but the safety guarantees derived rely on the accuracy of the numerical computation. Thus, it is crucial to understand and account for any inaccuracies that occur due to uncertainty in the underlying dynamics and environment as well as the induced numerical errors. To this end, we propose a framework for modeling the error of the value function inherent in Hamilton-Jacobi reachability using a Gaussian process. The derived safety controller can be used in conjuncture with arbitrary controllers to provide a safe hybrid control law. The marginal likelihood of the Gaussian process then provides a confidence metric used to determine switches between a least restrictive controller and a safety controller. We test both the prediction as well as the correction capabilities of the presented method in a classical pursuit-evasion example.

Keywords: Adaptive control, Autonomous systems, Intelligent systems
Abstract: This paper presents a balanced strategy for autonomous search problem from a control perspective, namely, dual control for exploration and exploitation (DCEE). To search an unknown source in an unknown environment, the agent is required to learn the operational environment and accomplish the control objective, which essentially forms a learning based control problem. A dual control for exploration and exploitation is developed to realise an optimal trade-off between reducing knowledge uncertainty and accomplishing required goal. Various algorithms in learning and control can be integrated into this new framework to offer flexible and customised solutions according to problem specifications and hardware performance. Relationships between DCEE and other algorithms, especially information-theoretic approaches and bio-inspired optimisation methods, are reflected from the perspective of exploration and exploitation. Simulation studies are provided to demonstrate the advantages of DCEE.

Keywords: Autonomous robots, Intelligent systems, Machine learning
Abstract: We present the design of a learning-based compliance controller for assembly operations for industrial robots. We propose a solution within the general setting of learning from demonstration (LfD), where a nominal trajectory is provided through demonstration by an expert teacher. This can be used to learn a suitable representation of the skill that can be generalized to novel positions of one of the parts involved in the assembly, for example the hole in a peg-in-hole (PiH) insertion task. Under the expectation that this novel position might not be entirely accurately estimated by a vision or other sensing system, the robot will need to further modify the generated trajectory in response to force readings measured by means of a force-torque (F/T) sensor mounted at the wrist of the robot or another suitable location. Under the assumption of constant velocity of traversing the reference trajectory during assembly, we propose a novel accommodation force controller that allows the robot to safely explore different contact configurations. The data collected using this controller is used to train a Gaussian process model to predict the misalignment in the position of the peg with respect to the target hole. We show that the proposed learning-based approach can correct various contact configurations caused by misalignment between the assembled parts in a PiH task, achieving high success rate during insertion. We show results using an industrial manipulator arm, and demonstrate that the proposed method can perform adaptive insertion using force feedback from the trained machine learning models.

Keywords: Machine learning, Communication networks, Optimization algorithms
Abstract: Modern cyber-physical architectures use data collected from systems at different physical locations to learn appropriate behaviors and adapt to uncertain environments. However, an important challenge arises as communication exchanges at the edge of networked systems are costly due to limited resources. This paper considers a setup where multiple agents need to communicate efficiently in order to jointly solve a reinforcement learning problem over time-series data collected in a distributed manner. This is posed as learning an approximate value function over a communication network. An algorithm for achieving communication efficiency is proposed, supported with theoretical guarantees, practical implementations, and numerical evaluations. The approach is based on the idea of communicating only when sufficiently informative data is collected.

Keywords: Machine learning, Lyapunov methods, Stability of nonlinear systems
Abstract: There has been a recent interest in imitation learning methods that are guaranteed to produce a stabilizing control law with respect to a known system. Work in this area has generally considered linear systems and controllers, for which stabilizing imitation learning takes the form of a biconvex optimization problem. In this paper it is demonstrated that the same methods developed for linear systems and controllers can be readily extended to polynomial systems and controllers using sum of squares techniques. A projected gradient descent algorithm and an alternating direction method of multipliers algorithm are proposed as heuristics for solving the stabilizing imitation learning problem, and their performance is illustrated through numerical experiments.

Keywords: Control education, Computer aided learning, Control courses and labs
Abstract: In recent years the IFAC technical committee has been looking closely at what constitutes an ideal first course in control cite{surveyjournal}. Having completed stage 1 of this study, the community is now discussing how to best collate and share the numerous open access resources to support such a course, or indeed more advanced control courses, which staff and students can utilise. A call has gone out to the community to share their knowledge of available resources and this paper gives a preliminary view on the resources that have been proposed so far and ideas on how this information will be shared with the community.

Keywords: Control education, Computer aided learning, Control courses and labs
Abstract: We propose a full receding-horizon approach to solve the problem of adaptively selecting the contents of a test while the test is being taken by a learner, while accounting for the past performance and the knowledge uptake needs of the learner itself. We thus present an approach that embeds a moving-horizon estimator to robustly infer the current knowledge levels of the learner, together with a model predictive controller to promote the robust selection of which items should be included in the test as it is being taken by the learner. Both receding-horizon algorithms are built on top of a model of the learners' knowledge uptake dynamics that accounts for Zone of Proximal Development effects, i.e., the hypothesis for which there exists an optimal difficulty level for the questions to maximize the learning at each step. We show that the proposed closed-loop approach outperforms the current policies that build tests either by randomly selecting questions within a database, or by making the difficulty of the questions non-decreasing as the test progresses.

Keywords: Control education, Control courses and labs, Mechatronics
Abstract: In this work we present a low-cost Arduino-based line follower robot designed to be used in basic control courses following a project-based-learning approach. The idea is that students can learn electronics, programming, signal processing and control through the construction and control of this mobile robot both in practical teaching classes in the laboratory and in their personal time outside school hours. The robot has been used in the second year basic control course of the Telecommunication Engineering degree of the University of Seville with over 180 students. The pedagogic objectives of this course and how it has affected the design are discussed. In addition, the design of the robot is presented in detail.

Keywords: Observers for nonlinear systems, Output regulation, Machine learning 
Abstract: Sixty years ago D. Luenberger published a landmark article entitled Observing the state of linear system in the journal IEEE Transaction on Military Electronics. It is in this article that the problem of asymptotic observer synthesis for linear systems is solved. This article had a big impact in the community and is now a classics of control theory. Since then, many attempts have been made to extend these techniques to the non-linear context. In this talk is given the general structure of a KKL observer. Secondly, we provide the theoretical tools that allow its study. We then show that the existence of such an algorithm is obtained under very weak hypotheses. After that, we make the link with other estimation algorithms. Finally, we show what are the key elements that allows its synthesis.

Keywords: Observers for nonlinear systems, Output regulation, Machine learning
Abstract: See Part 1.

Keywords: Observers for nonlinear systems, Output regulation, Machine learning
Abstract: See Part 1.

Keywords: Observers for nonlinear systems, Electrical machine control
Abstract: (Time-varying) KKL observers happen to be very well suited to solve observation problems in electrical motors. Indeed, such problems can often be reformulated as state estimation of a linear model with polynomial outputs, for which the design of the KKL observer is explicit, with a polynomial transformation whose coefficients depend on filtered versions of the currents/voltages. Applications include - estimation of rotor position and magnet flux/resistance for PMSMs, - estimation of rotor speed and torque for induction motors, from electrical measurements only (``sensorless'' setting). The advantage of the KKL design lies in the fact that it does not require the differentiation of the (noisy) currents/voltages and is able to provide information even in presence of indistinguishable trajectories (as in the position and resistance estimation problem for PMSMs).

Keywords: Observers for nonlinear systems, Output regulation
Abstract: Output regulation is a fundamental problem in control theory generalizing the problems of disturbance rejection and tracking in presence of uncertain parameters. Most existing approaches gravitate around the notion of internal model, a device able to generate the ideal steady-state control action rendering invariant a set where all the tracking and rejection objectives are met. The theory of nonlinear KKL observers proved to be very effective for the design of internal models, yielding the most general existence results, and removing long-standing immersion assumptions limiting the applicability of the existing designs. We cover the basic theoretical results pertaining the application of KKL observers to the design of regulators, and we overview some possible approximation methods and adaptive extensions providing additional robustness and learning capabilities.

Keywords: Observers for nonlinear systems, Output regulation, Machine learning
Abstract: A solution to implement nonlinear Luenberger observers is to approximate them by performing nonlinear regression on data simply generated by solving the system and observer dynamics. We detail different approaches for autonomous and excited systems and the choices made for data generation, pre-processing, and regression.

Keywords: Observers for nonlinear systems, Output regulation, Machine learning
Abstract: The problem of output prediction is the problem of designing a model for autonomous nonlinear systems capable of forecasting their future observations. To solve this problem it is possible to define a general framework bringing together the necessary properties for the development of such an output predictor. In particular, we look at this problem from two different viewpoints, control theory and data-driven techniques (machine learning), and try to formulate it in a consistent way, reducing the gap between the two fields. Building on this formulation and problem definition, we propose a predictor structure based on the KKL observer and we show that KKL fits well into our general framework. Finally, we propose a constructive solution for this predictor that solely relies on a small set of trajectories measured from the system. Our experiments show that our solution allows to obtain an efficient predictor over a subset of the observation space.

Keywords: Observers for nonlinear systems, Output regulation, Machine learning
Abstract: See Part 1.

Keywords: Nonlinear system theory, Stability of nonlinear systems, Optimization
Abstract: This session will review the history of graphical tools in systems and control, and address the question: why are graphical tools still relevant today?

1. Nyquist and Bode for uncertainty and robustness in linear control, and the historical connections with circuit theory. 2. The importance of incremental stability for nonlinear system analysis. 3. Graphical tools for nonlinear control: the circle and Popov criteria, connection to small gain and passivity theorems. 4. Incremental/differential analysis of nonlinear systems: state-of-the art.

Keywords: Nonlinear system theory, Stability of nonlinear systems, Optimization
Abstract: This session will introduce the mathematics of the Scaled Relative Graph, its use in optimization and the connection to classical control.

1. Definition, SRGs as a way to represent the incremental behavior of an operator on the complex plane. 2. Graphical algebra of SRGs, SRGs of operator properties. 3. SRGs in optimization, graphical proofs of convergence. 4. SRGs for nonlinear system analysis: SRGs of LTI transfer functions, static nonlinearities and their compositions.

Keywords: Nonlinear system theory, Stability of nonlinear systems, Optimization
Abstract: See Part 1.

Keywords: Nonlinear system theory, Stability of nonlinear systems, Optimization
Abstract: See Part 1.

Keywords: Nonlinear system theory, Stability of nonlinear systems, Optimization
Abstract: This session will explore the use of SRGs for the study of nonlinear feedback systems, and in particular, their incremental stability.

1. Graphical proofs of the small gain and passivity theorems. 2. A general stability theorem for feedback interconnection. 3. Rolled-off passivity. 4. Examples.

Keywords: Nonlinear system theory, Stability of nonlinear systems, Optimization
Abstract: See Part 1.

Keywords: Nonlinear system theory, Stability of nonlinear systems, Optimization
Abstract: See Part 1.

Keywords: Nonlinear system theory, Stability of nonlinear systems, Optimization
Abstract: We will conclude with some open questions which are the subject of ongoing research, including characterising the SRGs of switched systems, nonlinear loop shaping and estimating SRGs from sampled data.

Keywords: Game theoretical methods
Abstract: How should a system operator strategically allocate its assets in an adversarial environment? In this talk, we explore this question in the context of the well-studied Colonel Blotto game. Colonel Blotto games model strategic scenarios where two opposing entities are tasked with allocating a given number of assets over a collection of fronts. The classical setting has primarily focused on scenarios where each front is associated with an independent valuation and each entity seeks to maximize the cumulative performance over the fronts. This talk will focus on emerging variants of these Colonel Blotto games encompassing both informational asymmetries and alternative structures for the classic linear sum objectives. One such variant we will focus on is a “networked” Colonel Blotto game, where the valuation of a front now depends on the outcome associated with other fronts in the system.

Keywords: Energy systems, Power plants
Abstract: Though wave energy systems are not yet commercial, control has been identified as an important enabling technology which can reduce the cost of wave energy, allowing it to compete economically with other renewable and conventional energy sources. However, wave energy systems, which are diverse in form and operating principle, represent a challenging control problem, in terms of panchromatic reciprocating energy flux, hydrodynamic modelling complexity, non-causality in the fundamental control solution, and adverse sensitivity properties. In addition, the wave energy control problem is expressed in terms of an energy maximising performance function, rather than being easily reduced to a set-point following problem, while a key system input variable, the wave excitation force, is unmeasurable.

This talk will detail the major control issues faced in dealing with wave energy systems, also providing an overview of wave energy technology and some typical devices, while showing some possibilities in the solution domain. Some experimental control results will also be presented and the talk will conclude with some perspectives on future research directions.

Keywords: Transportation systems, Reduced order modeling, Stochastic filtering
Abstract: State estimation for traffic networks is a particularly challenging problem in view of their large dimensionality, and since models are often inaccurate and the interaction patterns unpredictable. In this article, we approach the problem by mixing aggregation-based complexity reduction and nonlinear filtering. We subdivide vehicles into groups and derive a lower-dimensional approximate model where vehicles belonging to the same group are represented by a unique random variable matching their average characteristics. Then, we propose a procedure to estimate the statistical properties of the group variables from partial measurements. Connections to car-following models are discussed, and the developed methodology is illustrated through numerical simulations.

Keywords: Traffic control, Transportation systems
Abstract: Bi-modal delivery systems are a promising solution to the challenges posed by the increasing demand of e-commerce. Due to the potential benefit drones can have on logistics networks such as delivery systems, some countries have taken steps towards integrating drones into their airspace. In this paper we aim to quantify this potential by developing a mathematical model for a Bi-modal delivery system composed of trucks and drones. We propose an optimization formulation that can be efficiently solved in order to design socially-optimal routing and allocation policies. We incorporate both societal cost in terms of road congestion and parcel delivery latency in our formulation. Our model is able to quantify the effect drones have on mitigating road congestion, and can solve for the path routing needed to minimize the chosen objective. To accurately capture the effect of stopping trucks on road latency, we model it using SUMO by simulating roads shared by trucks and cars. Based on this, we show that the proposed framework is computationally feasible to scale due to its reliance on convex quadratic optimization techniques.

Keywords: Transportation systems, Modeling, Optimization
Abstract: In this paper, a day-to-day discrete-time traffic assignment model is introduced to represent the evolution of users' choices and their progressive adjustments towards new equilibria in transport networks in which new conditions occur. Specifically, the proposed model considers a proportional switch assignment process, in which every day the users decide whether to maintain the same path chosen the previous day or to switch it, not only on the basis of the congestion level experienced on the chosen routes but also according to the extent of topological similarity between the potential new paths and the one already used. In this paper, the potential of this model to analyse the performance degradation in a transport network affected by disruptive events is shown by reporting two simulation examples.

Keywords: Traffic control, Game theoretical methods, Optimization
Abstract: We propose an incentive-based traffic demand management policy to alleviate traffic congestion on a road stretch that creates a bottleneck for the commuters. The incentive targets electric vehicles owners by proposing a discount on the energy price they use to charge their vehicles if they are flexible in their departure time. We show that, with a sufficient monetary budget, it is possible to completely eliminate the traffic congestion and we compute the optimal discount. We analyse also the case of limited budget, when the congestion cannot be completely eliminated. We compute analytically the policy minimising the congestion and estimate the level of inefficiency for different budgets. We corroborate our theoretical findings with numerical simulations that allow us to highlight the power of the proposed method in providing practical advice for the design of policies.

Keywords: Transportation systems, Autonomous systems, Automotive
Abstract: Control design for safe and time-efficient motion of automated vehicles in roundabout scenarios poses various challenges, especially adaptation to the actual traffic scenario and coordination of the vehicles. This paper proposes the design of a hierarchical motion control with learning functionality for roundabout scenarios. The control is designed on two levels, such as on cloud level and on vehicle level. The control on the cloud level is designed by using reinforcement learning, with which the maximum speed for the vehicle is achieved. The vehicle level contains a robust controller and a supervisor, with which the collision avoidance of the vehicles is guaranteed. The proposed control on Hardware-in-the-Loop environment with small-scaled indoor vehicles is implemented. The effectiveness of the control and the safe motion of the automated vehicle under various scenarios are demonstrated. The provided scenarios illustrate that safe, i.e., collision-free motion of the automated vehicle can be guaranteed, even if the connection to the cloud has been lost.

Keywords: Transportation systems, Coverage control, Distributed control
Abstract: Ride-sourcing system can provide passengers with fast and efficient service with a fleet of vehicles, while asymmetry between origin and destination distributions of trips, nonuniform passenger's demand for rides in different districts creates imbalances in the spatial distribution of these vehicles. Thus proactively relocating idle vehicles to the high-demand regions, also known as vehicle rebalancing is an emerging problem that can have a significant improvement for the efficiency of urban transportation. We formulate this problem as a coverage problem for coordination and deployment of multiple mobile agents in city scenarios, which vehicles can benefit from by allocating them according to the different demand densities of different city districts. A Voronoi-based control algorithm is proposed by leveraging the local information of each vehicle. The effectiveness of the proposed method is validated by a simulator modeled on a real road map from Shenzhen, China. Compared to baseline, our proposed method is able to serve more trips with less passenger waiting time.

Keywords: Transportation systems, Game theoretical methods, Decentralized control
Abstract: Both ride-hailing services and electric vehicles are becoming increasingly popular and it is likely that charging management of the ride-hailing vehicles will be a significant part of the ride-hailing company's operation in the near future. Motivated by this, we propose a game theoretic model for charging management, where we assume that it is the fleet-operator that wants to minimize its operational cost, which among others include the price of charging. To avoid overcrowded charging stations, a central authority will design pricing policies to incentivize the vehicles to spread out among the charging stations, in a setting where several ride-hailing companies compete about the resources. We show that it is possible to construct pricing policies that make the Nash-equilibrium between the companies follow the central authority's target value when the desired load is feasible. Moreover, we provide a decentralized algorithm for computation of the equilibrium and conclude the paper with a numerical example illustrating the results.

Keywords: Optimization, Autonomous robots, Maritime
Abstract: The performance of a guidance, navigation and control (GNC) system of an autonomous underwater vehicle (AUV) heavily depends on the correct tuning of its parameters. Our objective is to automatically tune these parameters with respect to arbitrary high-level control objectives within different operational scenarios. In contrast to literature, an overall tuning is performed for the entire GNC system, which is new in the context of autonomous underwater vehicles. The main challenges in solving the optimization problem are computationally expensive objective function evaluations, crashing simulations due to infeasible parametrization and the numerous tunable parameters (in our case 13). These challenges are met by using constrained Bayesian optimization with crash constraints. The method is demonstrated in simulation on a GNC system of an underactuated miniature AUV designed within the TRIPLE-nanoAUV initiative for exploration of subglacial lakes. We quantify the substantial reduction in energy consumption achieved by tuning the overall system. Furthermore, different parametrizations are automatically generated for different power consumption functions, robustness, and accuracy requirements. E.g. energy consumption can be reduced by ~28%, if the maximum allowed deviation from the planned path is increased by ~65%. This shows the versatile practical applicability of the optimization-based tuning approach.

Keywords: Optimization, Control over networks, Autonomous systems
Abstract: In this manuscript, we investigate the design-for-control problem to optimize locations and operational settings of chlorine boosters in water networks. The objective is to minimize deviations from target chlorine concentrations. The problem formulation includes discretized linear PDEs modelling advective transport of chlorine concentrations. Moreover, binary variables model the placement of chlorine boosters. The resulting optimization problem is a convex mixed integer program (MIP), which is difficult to solve, especially when large water networks are considered. We develop a new swapping heuristic to optimally place and control chlorine boosters in water networks. The proposed method relies on a continuous relaxation of the original MIP. We evaluate the heuristic using two case studies, including one large operational water network from the UK.

Keywords: Optimization, Distributed estimation over sensor nets, Distributed cooperative control over networks
Abstract: This paper considers distributed convex optimization problems with nonnegativity constraints on the decision variables. We propose a novel distributed gradient dynamics in continuous-time setting that can solve this problem. In contrast to the existing methods in the literature, we do not incorporate any Euclidean projection operators or penalty functions to obtain an approximate solution. We show that the proposed network flow is guaranteed to converge asymptotically, on any connected graph, to the unique global minimizer, provided that the aggregate objective function is strongly convex and the local cost functions have Lipschitz-continuous gradients.

Keywords: Optimization, Identification, Modeling
Abstract: In this contribution we consider sparse linear regression problems. It is well known that the mutual coherence, i.e. the maximum correlation of the regressors, is important for the ability of any algorithm to recover the sparsity pattern of an unknown parameter vector from data. A low mutual coherence improves the ability of recovery. In optimal experiment design this requirement may be in conflict with other objectives encoded by the desired Fisher matrix. In this contribution we alleviate this issue by combining optimal input design with a recently proposed approach to achieve low mutual coherence by way of a linear coordinate transformation. The resulting optimization problem is solved using cyclic minimization. Via simulations we demonstrate that the resulting algorithm is able to achieve a Fisher matrix which results in a performance close to the performance if the sparsity would have been known, while at the same time being able to recover the sparsity pattern.

Keywords: Optimization, Predictive control for nonlinear systems, Robotics
Abstract: A monolithic integration of the robot online task allocation and trajectory planning within the framework of a hybrid model predictive controller is introduced. To this end, the underlying mixed-integer nonlinear programming (MINLP) problem is transformed into a relaxed mixed-integer quadratically constrained programming (MIQCP) problem, suitable to generate feasible robot trajectories online. The proposed algorithm is implemented and validated on an experimental setup using the robot operating system (ROS) software and a robot arm performing discrete pick-and-place tasks.

Keywords: Optimization, Transportation systems, Maritime
Abstract: One of the main problems of inland waterways is the scheduling and trip planning of vessels, which is even more challenging when the waterway is a natural channel or river whose depth and width are conditioned by its bathymetric profile, the effect of the tide and the orography of the region. This is the case of the Guadalquivir river connecting the Atlantic Ocean and the inland port of Seville in Spain. The correct scheduling and trip planning of vessels is not only important to ensure security, but it also plays an essential role in the efficiency and prestige of the port. To address this problem we propose the formulation of a mixed-integer linear program that aims to reduce the time vessels spend waiting and sailing through the waterway while taking into account depth and width constraints to ensure the safety of crossing and overtaking manoeuvres. The result is an optimal plan that indicates the times in which the vessels should pass through a sequence of predefined waypoints of the waterway. To demonstrate the benefits of the proposed method, a set of simulations have been carried out. The results have been analysed in terms of optimality and computational complexity in comparison to a first-arrived first-served scheduling method.

Keywords: Process control, Neural networks, Optimization
Abstract: This paper proposes a method for lifelong learning of Recurrent Neural Networks, such as NNARX, ESN, LSTM, and GRU, to be used as plant models in control system synthesis. The problem is significant because in many practical applications it is required to adapt the model when new information is available and/or the system undergoes changes, without the need to store an increasing amount of data as time proceeds. Indeed, in this context, many problems arise, such as the well known Catastrophic Forgetting and Capacity Saturation ones. We propose an adaptation algorithm inspired by Moving Horizon Estimators, deriving conditions for its convergence. The described method is applied to a simulated chemical plant, already adopted as a challenging benchmark in the existing literature. The main results achieved are discussed.

Keywords: Observers for nonlinear systems, Autonomous robots, Maritime
Abstract: This work presents a set-membership state estimator for autonomous surface vehicles, based on an augmented state including lumped disturbances. The position and orientation are assumed to be measured subject to bounded noises. A novel dynamical decomposition decouples the estimation problem into two simpler subproblems, for the rotational and positional dynamics. Then, under physically motivated assumptions about the vessel maximum velocities and acceleration rates, the estimator computes sets enclosing the positions, velocities, and lumped generalised disturbances gathering several kinds of modelling uncertainties. The sets are described by zonotopes. A set-based estimation of the lumped generalised disturbances paves the way toward an enhanced motion control scheme, where low-level controllers could compensate them, depending on the estimation accuracy. Several simulations with a well-known test-bed craft compare the performance of the proposed algorithm with a previous one from the literature under realistic environmental conditions.

Keywords: Autonomous robots, Predictive control for linear systems, Cooperative autonomous systems
Abstract: This paper addresses a multi-robot task allocation (MRTA) towards moving tasks and presents a novel computationally efficient predictive allocation algorithm that requires solving a linear program (LP) problem. Following the receding horizon control policy, the present algorithm repeats the optimization of future task assignments within an emph{allocation horizon} while predicting the evolution of the system. The online optimization is formulated so that the assignment problem is reduced exactly to an LP. The algorithm is also compared with other traditional methods, namely, the greedy approach and a genetic algorithm (GA). Our results show that the algorithm here proposed outperforms the greedy approach for small prediction horizons and has significantly lower computational load than GA.

Keywords: Autonomous robots, Transportation systems
Abstract: This paper presents a novel lateral controller that solves the path tracking problem. The proposed controller uses an orthogonal projection on a prescribed path to stabilise an error model, which leads to constructive conditions on the controller parameters. Explicit formulas to determine the orthogonal projection for straight and circular paths are given and it will be shown that the mobile system stays on an appropriately designed path once it reaches the path. The theoretical results are verified by laboratory experiments.

Keywords: Automotive, Autonomous robots, Linear parameter-varying systems
Abstract: This paper presents an approach for state estimation of an autonomous vehicle using a Linear Parameter Varying Kalman Filter and SLAM. The proposed state estimation schema is integrated with an MPC controller in charge of controlling the vehicle and should be able to estimate the vehicle state in presence of noises and system disturbances. The proposed approach uses the SLAM technique to obtain the rough pose estimation of the vehicle. Furthermore, using the inaccurately estimated pose from SLAM and measurement from other sensors, the full state of the vehicle is estimated and the map is corrected using Linear Parameter Varying modeling approach and online gain interpolation. The optimal and stable gain is obtained by formulating the design conditions using LMIs and by exploiting Lyapunov stability criteria, and dual LQR optimal design. The proposed approach is tested in the simulation and for the practical validity using a small-scale autonomous car.

Keywords: Stochastic systems, Autonomous robots, Robotics
Abstract: We present an analytical method to estimate the continuous-time collision probability of motion plans for autonomous agents with linear controlled Ito dynamics. Motion plans generated by planning algorithms cannot be perfectly executed by autonomous agents in reality due to the inherent uncertainties in the real world. Estimating end-to-end risk is crucial to characterize the safety of trajectories and plan risk optimal trajectories. In this paper, we derive upper bounds for the continuous-time risk in stochastic robot navigation using the properties of Brownian motion as well as Boole and Hunter's inequalities from probability theory. Using a ground robot navigation example, we numerically demonstrate that our method is considerably faster than the naive Monte Carlo sampling method and the proposed bounds perform better than the discrete-time risk bounds.

Keywords: UAV's, Autonomous robots
Abstract: This paper develops a novel learning-based local path planning method for Unmanned Aerial Vehicles (UAVs) in unknown environments. We establish a neural network (NN) with two fully connected hidden layers, where the distances from the UAV to the hit points of the locally detected obstacles, a strategic temporary goal and the direction to the final destination are selected as the input of the NN, and the reference velocity for the UAV to track is chosen as the output. To collect the training data, we propose a local path planning method, which repeatedly constructs a local Laplacian Potential Field (LPF) only based on the UAV's real-time obstacle detections of limited scope, and requires the UAV to track the negative gradient direction of the resulting potential function. Then, the UAV follows the reference velocity generated by the trained NN path planner to safely approach the final destination. Simulations demonstrate the effectiveness, adaptability, and efficiency of the proposed learning-based path planning method, which outperforms the above LPF-based path planning method and, unlike many other learning-based methods, does not need to re-train the NN parameters when changed to new maps.

Keywords: H2/H-infinity methods, Autonomous robots, Robotics
Abstract: This paper proposes an optimized lateral control design of a two-steering-axle off-road vehicle. An extension of the well-known bicycle model is introduced that considers two steering axles and the slope of the local ground. Lateral deviation controllers are thereby synthesized to quantitatively evaluate the use of multiple steering angles. The proposed multi-input multi-output feedback controllers originate from a H2/Hinf multi-objective synthesis, achieving optimum trade-offs between performance and robustness. Three controller architectures are considered (depending on actuation possibilities) and compared qualitatively and quantitatively. In addition, realistic simulation results are analyzed using a non-linear simulator that considers the fine modeling of sensors and actuators (e.g., hydraulic actuator). The results are promising and will be the subject of an experimental validation project.

Keywords: Delay systems, Stability of linear systems, Uncertain systems
Abstract: This paper addresses the stability analysis of uncertain discrete-time delayed systems described by Difference-Algebraic Representations (DAR). This kind of representation allows dealing with rational functions of the state as well as uncertain parameters. New delay-dependent Linear Matrix Inequalities based-conditions are obtained with a parameter-dependent Lyapunov-Krasovskii functional jointly with the Wirtinger-based summation inequality and the delay-dependent Moon's inequality with a suitable augmented vector. Numerical examples illustrate the potential and effectiveness of the proposed conditions.

Keywords: Power electronics, Stability of nonlinear systems, Electrical power systems
Abstract: We consider a grid-connected voltage source converter (VSC) and address the problem of estimating the grid angle and frequency---information that is essential for an appropriate operation of the converter. We design phase-locked loop (PLL) algorithms with guaranteed stability properties, assuming that the grid is characterized by relatively high short-circuit-ratio and inertia. In particular we derive, using passivity arguments, generalization of the conventional synchronous reference frame and arctangent PLL, which are the standard solutions in power applications. The analysis is further extended to the case of interconnection to a low-inertia grid, ensuring robustness of the algorithms in case of frequency variations. The results are validated on a benchmark including the grid, the VSC and related controllers.

Keywords: Stability of nonlinear systems, LMI's/BMI's/SOS's, Optimal control
Abstract: In this paper we focus on nonlinear systems composed by a linear term plus a nonlinearity satisfying a monotonic or a sector bound condition. We present sufficient conditions based on LMIs for the design of a feedback control law possessing an infinite gain margin property that makes the closed-loop system to define a contraction. Then, we analyze the connection between infinite gain margin contractive feedbacks and optimization problems by showing how such an infinite gain control design minimizes a given cost function. A practical case of study is given to illustrate our results.

Keywords: Control over communication, Stability of linear systems, Statistical learning
Abstract: In the analysis and control of discrete-time linear time-invariant systems, the spectral radius of the system state matrix plays an essential role. Usually, it is assumed that system matrices are known, from which the spectral radius can be directly computed. Instead, we consider the setting where the system is affected by process noise, and one has only finitely many samples of system input and state measurements. We provide two methods for estimating the spectral radius and derive error bounds that hold with high probability. Moreover, we show how to use the derived results to test stabilizability for networked control systems (NCSs) with lossy channels when only finitely many samples of the system input, state, and packet drop sequence are available.

Keywords: Control over networks, Genetic regulatory systems, Stability of linear systems
Abstract: For the realization of ultra-early medical care, a method to detect early-warning signals of the transitions to a disease state, which are considered as critical transitions (bifurcation) of dynamical systems, has been developed since 2014. Towards developing a new treatment at an ultra-early medical stage when a critical transition just occurs, in this paper, we address a stabilization problem of an undirected network system by the pole placement method, where no network system model is available, from the control engineering point of view, and propose a theoretical method to design the optimal single-input assignment with high-dimension, small-sample-size data. In addition, we propose an approximate design method based on observed data and show its effectiveness by numerical simulations.

Keywords: Distributed parameter systems, Sampled data control, Stability of nonlinear systems
Abstract: This paper presents the design of an event-triggering mechanism for the damped linear Schrödinger equation. Localized damping is considered. The absence of any accumulation points of the time updates sequence is proven, ensuring the avoidance of Zeno behavior. The global exponential stability is ensured through some energy estimates exploiting observability inequality. An illustrative example based on the one dimensional Schrödinger equation demonstrates the efficiency of the results.

Keywords: Lyapunov methods, Robust control, Stability of nonlinear systems
Abstract: The current paper proposes a cascade structure controller for use with finite-order input-affine nonlinear systems with polytopic linear time-invariant (LTI) bounds. The controller is comprised of a Krasovskii-passivity based component, designed for polytopic systems, which guarantees closed-loop asymptotic stability of the nonlinear plant irrespective of the operating point, followed by a robust path-planning component for transient and steady-state performance, designed through mu-synthesis. A step-by-step design methodology, along with the performance of the proposed approach are illustrated using a benchmark uncertain plant with nonlinear dynamics.

Keywords: Traffic control, Transportation systems, Modeling
Abstract: In this extended abstract, we propose a novel model that describes how the traffic evolution on a highway stretch is affected by the presence of a service station. The presented model enhances the classical Cell Transmission Model (CTM) dynamics by adding the dynamics associated to the service stations, where the vehicles may stop before merging back in the mainstream. We call this novel model Cell Transmission Model with service station (CTM-s) and we discuss its ability to describe different complex scenarios in which multiple service stations are present.

Keywords: Energy systems, Optimal control, Reduced order modeling
Abstract: This work investigates the use of LQ-optimal state feedback controllers for large floating wind turbines using the recent 15MW UMaine VolturnUS-S semi-submersible floating model, comparing an LQG controller with integral action to the `ROSCO' PI controller used as a standard baseline. Similar LQ-optimal controllers already showed promising results on smaller turbines, but their performance has not yet been investigated on this 15MW turbine. In addition, we consider the use of a Kalman filter to estimate the turbine state, which is often omitted in similar works, for a full LQG design. Testing was performed on a nonlinear model using OpenFAST. Results show significantly lower generator speed, platform pitch and tower top fluctuations and deflections, with reduced settling times of more than a 100s compared to the ROSCO controller, confirming the efficacy of LQG controllers for the 15MW floating turbine. This enables future work to investigate the performance under turbulent wind conditions and develop specific controller improvements for large turbines.

Keywords: Game theoretical methods, Control over networks, Decentralized control
Abstract: We consider linear-quadratic-Gaussian (LQG) games in which players have quadratic payoffs that depend on the players' actions and an unknown payoff-relevant state, and signals on the state that follow a Gaussian distribution conditional on the state realization. An information designer decides the fidelity of information revealed to the players in order to maximize the social welfare of the players. Leveraging the semi-definiteness of the information design problem, we derive analytical solutions for welfare maximization under specific LQG games. We show that full information disclosure maximizes social welfare when there is a common payoff-relevant state, when there is specific strategic substitutability in the actions of players, or when the signals are public. Numerical results show that as strategic substitution increases, the value of the information disclosure increases.

Keywords: Game theoretical methods, Decentralized control, Agents networks
Abstract: We analyze convergence of decentralized fictitious play (DFP) in near-potential games, where agents participate in a game in which the change in utility functions are closely aligned with a potential function. In DFP, agents take actions that maximize their expected utilities computed based on local estimates of empirical frequencies of other agents. These local estimates are updated by averaging estimates received from neighbors in a time-varying communication network. Given near-potential games with finitely many Nash equilibria that are distant enough from each other, we show that the empirical frequencies converge near a single Nash Equilibrium. This result establishes that DFP maintains the properties of standard fictitious play (FP) in near-potential games.

Keywords: Iterative learning control, Adaptive control, Lyapunov methods
Abstract: Repetitive processes make repeated applications, termed passes or trials, through a set of dynamics defined over a finite duration, known as the trail length. These processes can be used to model industrial examples such as long-wall coal cutting and metal rolling. The distinguishing feature is that the output on any trial acts as a forcing function on the next trial and therefore contributes to its dynamics, and the result can be oscillations that increase in amplitude from trial-to-trial. Moreover, this unacceptable feature cannot be regulated by standard control action and instead stability theory and control law design must be approached in the 2D systems setting, i.e., information propagation from trial-to-trial and along the trials.

In the case of linear dynamics, a stability and control theory has been developed based on representing the dynamics by a bounded linear operator in a Banach space setting. This has then led on to control law design with experimental validation in the area of iterative learning control. An immediate question then is: can this progress be replicated for nonlinear dynamics? The answer to this question can be for a full nonlinear model of the dynamics or for cases where nonlinearity is introduced by nonlinear actuator dynamics.

This contribution explains how a stability theory for nonlinear repetitive processes can developed using vector Lyapunov functions, followed by extension to passivity based control law design. Next, iterative learning control law design for systems with saturating actuators is developed from this theory. The analysis covers both deterministic and stochastic designs and supporting experimental results from a servo motor based example are given and discussed.

Keywords: LMI's/BMI's/SOS's, Electrical power systems, Network analysis and control
Abstract: This paper is concerned with the feasibility of the power flow in DC power power grids with constant power loads. Necessary and sufficient matrix inequalities are derived that guarantee a minimal p-norm distance between a configuration of power demands and the unfeasibility boundary in the space of power demands. The (non)convexity of these matrix inequalities is studied subsequently.

Keywords: Markov processes, Modeling, Linear systems
Abstract: Random walks have been frequently used by researchers to investigate epidemic networks, with the ultimate goal of preventing, or at least mitigating, the disease spread. However, models based on random walks only approximate the true dynamics of an epidemic spreading, and it is not always clear whether the predictions of such approximated models are eventually reliable or not. In this short abstract, we first show that such models may provide largely inaccurate predictions when compared with simulated outcomes, and then we show how alternative Markov chain models may be constructed to better reflect the dynamics of the virus (i.e., in terms of the mean time to absorption).

Keywords: Mechatronics, Lyapunov methods, Nonlinear system theory
Abstract: This extended abstract describes methodologies to tune the parameters of a class of passivity-based controllers for standard nonlinear mechanical systems. In particular, we provide tuning rules to ensure the desired stability margin, the upper bound on the rate of convergence, and the upper bound on the maximum permissible overshoot of the closed-loop. We conclude this extended abstract with experimental results that illustrate the applicability of the tuning rules in underactuated mechanical systems.

Keywords: Nonlinear system identification, Optimal control, Game theoretical methods
Abstract: The extended abstract addresses the problem of inverse learning for the linear-quadratic zero-sum differential games, where the aim is to evaluate the cost functions which lead to the observed stationary linear feedback law that constitutes the Nash equilibrium (NE) pair. We provide an algorithm that generates an equivalent game that shares one of the NE feedback laws with the original game. The algorithm combines both inverse optimal control and reinforcement learning methods in the form of gradient descent updates. We show the convergence of the algorithm and illustrate the algorithm’s performance in simulations.

Keywords: Adaptive control, Maritime, Identification for control
Abstract: This paper presents optimal sailing maneuvers for an AC75 foiling sailboat competing in the America’s Cup. The innovative sailboat design introduces extra degrees of freedom and articulations in the boat that result in nonlinear, high-dimensional, and unstable dynamics. The optimal maneuvers were achieved via the exploration of out-of-the-box solutions through adaptive control and optimization. We used a high-fidelity sailboat simulator for the data generation process, and an adaptive control approach (Jacobian Learning (JL)) to optimize the sailing maneuvers. These maneuvers serve as benchmarks and provide insightful information about the underlying dynamics of the boat. The close-hauled and tacking maneuvers were optimized to achieve maximum Velocity Made Good (VMG) and minimum loss of VMG, respectively. The optimal solutions are subject to physical/actuator constraints as well as the ones enforced to ensure the feasibility of the maneuvers by humans (sailors). The optimal maneuvers boast a marginal loss in sailing performance (VMG) of less than 1.5%, which enables exploiting areas of good wind conditions in the racing environment by maneuvering towards these areas without accruing the significant losses traditionally associated with performing multiple maneuvers.

Keywords: Aerospace, Automotive, Neural networks
Abstract: Due to safety reasons, anti-skid braking control strategies in aircraft can rely only on sensors that are integral to the landing gear, and for this reason they are currently developed as threshold-based algorithms that regulate the wheel deceleration. However, as it is known from the Automotive field, slip control offers superior closed-loop properties and robustness, paired with a much easier and time-saving tuning phase. This work aims at investigating first of all if slip control does indeed offer the same performance advancement also in the aeronautical braking context and, secondly, whether the wheel slip can be effectively estimated and employed in a closed-loop braking controller without the need of additional sensors on the landing gear. To do this, we propose a wheel slip control scheme where a data-driven approach using a neural network architecture is used to solve the problem of wheel slip estimation. The proposed control scheme is tested within a very realistic simulation setting, and compared with a standard deceleration-based ABS, showing superior performance and robustness.

Keywords: Transportation systems, Game theoretical methods
Abstract: This paper proposes to use a two-stage congestion game to model the decision making processes of passengers in a transportation system that incorporates heterogeneous transportation services. The structure of this game is analyzed to provide necessary and sufficient conditions to check whether a strategy profile in the game is a subgame perfect Nash equilibrium. We also quantified the inefficiency of a subclass of two-stage congestion games by showing that the lower bound of its sequential price of anarchy depends quadratically on the number of agents. Finally, we construct a possible model of a practical situation that achieves the quadratic dependence.

Keywords: Transportation systems, Modeling, Observers for nonlinear systems
Abstract: This study presents a simulation model of a locomotive which includes the longitudinal, vertical, pitch and wheelset rotational dynamics. The model captures the dynamics that are most important for wheel slip control. An estimation model is developed based on this simulation model. Initial simulations suggest that accurate and fast estimation of the adhesion forces of the individual wheelsets, as well as other important parameters including the coupler force and creep, can be achieved.

Keywords: Transportation systems, Optimization, Energy systems
Abstract: We optimize the operation of a fuel cell hybrid train using convex optimization. The main objective is to minimize hydrogen fuel consumption while achieving target journey duration. This is achieved by jointly optimizing train speed and the energy management system. Moreover, a battery thermal model and thermal constraints are integrated into the optimization problem to improve battery thermal management. The proposed joint optimization approach outperforms a conventional driving strategy both in terms of fuel consumption and battery thermal management. The joint approach consumes around less 10% fuel because its speed profile exploits kinetic energy recovery more effectively. Moreover, the joint approach requires 12% less cooling because its acceleration profile leads to less battery heating. Lastly, it was noticed that the joint approach scheduled battery cooling commands exclusively around periods of regenerative braking. This form of auxiliary load scheduling is believed to lead to marginally improved efficiency and battery thermal management.

Keywords: Constrained control, Neural networks, Automotive
Abstract: In this paper the application of Virtual Reference Feedback Tuning (VRFT) for control of nonlinear systems with regulators defined by Echo State Networks (ESNs) and Long Short Term Memory (LSTM) networks is investigated. The capability of this class of regulators of constraining the control variable is pointed out and a control scheme that allows to achieve zero steady-state error is presented. The developed scheme is validated on a benchmark example that consists of an Electronic Throttle Body (ETB).

Keywords: Constrained control, Optimization algorithms, Aerospace
Abstract: For control systems whose uncontrolled solutions are periodic (or more generally recurrent), there are geometric tools, developed in the 1980s, that assess controllability based on a Lie algebra rank condition, under the assumption that the control set contains zero in its interior. Motivated by solar sails control, the present study explores the case where zero is rather on the boundary of the control set. More precisely, it investigates the controllability of fast-oscillating dynamical systems subject to positivity constraints on the control variable, i.e., the control set is contained in a cone with vertex at the origin. A novel sufficient controllability condition is stated, and a constructive methodology is offered to check this condition, and to generate the controls, with values in the convex cone, that move, at first order, the slow state to an arbitrary direction of the tangent space. Controllability of a solar sail in orbit about a planet is analysed to illustrate the developments. It is shown that, given an initial orbit, a minimum cone angle parametrising the control set exists which satisfies the sufficient condition.

Keywords: Intelligent systems, Autonomous systems, Transportation systems
Abstract: The Vehicle routing problem (VRP) is widely discussed in the literature. The goal of the VRP is to provide an optimized traversal of a graph. This paper presents a mixed graph traversal optimization for the VRP. The mixed graph is a graph that contains both directed and undirected edges. The problem to be solved is a parking space search problem in a mixed graph where, from an initial location a vehicle first navigates to a zone, where the selected edges are located, and plans the traversal driving through all selected edges, and returning to the start of the exploration so that the path is repeatable. The paper shows, that the parking space search problem in a mixed graph can be reduced to a travelling salesman problem (TSP), for which an ant colony optimization (ACO) is implemented, to provide a solution, that is minimizing the total cost of the mixed graph traversal.

Keywords: Game theoretical methods, Agents and autonomous systems, Statistical learning
Abstract: We present a novel framework for a comparative analysis of stochastic decision strategies. In particular, we compare Stochastic Best Response (SBS) and Stochastic Better Response (SBT), which are important and widely adopted strategies in population games comprising players with bounded rationality. We propose an evolutionary game setting in which players repeatedly interact with each other and make decisions at two-time scales. At a lower level (faster time-scale) the decision set of each player is a finite action set and players select their actions to improve their payoffs in response to the action history of other players according to their stochastic learning strategies. At a higher level (slower time-scale), the action set of each player is a finite set of stochastic learning strategies and the players update their strategies according to imitation-mutation dynamics. To demonstrate our framework, we consider Prisoner’s Dilemma (PD) and Hawk Dove (HD), which are two of the most important 2 times 2 symmetric matrix games. For these games, we compare SBS and SBT under our proposed framework and establish that SBS dominates SBT in PD but SBT dominates SBS in HD.

Keywords: Game theoretical methods, Optimization
Abstract: This paper investigates the problem of generalized Nash equilibrium (GNE) seeking in population games under the Brown-von Neumann-Nash dynamics and subject to general affine equality constraints. In particular, we consider that the payoffs perceived by the decision-making agents are provided by a so-called payoff dynamics model (PDM), and we show that an appropriate PDM effectively steers the agents to a GNE. More formally, using Lyapunov stability theory, we provide sufficient conditions to guarantee the asymptotic stability of the set of generalized Nash equilibria of the game, for the case when the game is a so-called stable game (also known as contractive game). Furthermore, we illustrate the application of the considered framework to an energy market game considering coupled equality constraints over the players decisions.

Keywords: Game theoretical methods, Optimal control, Intelligent systems
Abstract: In this paper, we consider the problem of pursuit-evasion between two non-holonomic agents. To make the problem more realistic, we assume that the pursuer and evader have incomplete information regarding each other's motion. We propose a novel, two-phase, evasive strategy against a higher speed pursuer, based on worst-case scenario planning and proximity-based maneuver. In the first phase, the evader assumes the pursuer to have zero turn radius and executes a best-response strategy, which is analytically shown to be movement along a tangent to the turn circle of the evader. In the second phase, when the pursuer gets close to the evader, the latter switches to movement along high-curvature paths to side-step the former. The effectiveness of the two phases is shown through simulations.

Keywords: Game theoretical methods, Optimization algorithms, Variational methods
Abstract: We present a distributed Nash equilibrium seeking method based on the Bregman forward-backward splitting, which allows us to have a mirror mapping instead of the standard projection as the backward operator. Our main technical contribution is to show convergence to a Nash equilibrium when the game has cocoercive pseudogradient mapping. Furthermore, when the feasible sets of the agents are simplices, a suitable choice of Legendre function results in an exponentiated pseudogradient method, which, in our numerical experience, performs better than the standard projected pseudogradient and dual averaging methods.

Keywords: Machine learning, Game theoretical methods, Optimization
Abstract: We consider quadratic, nonmonotone generalized Nash equilibrium problems with symmetric interactions among the agents. Albeit this class of games is known to admit a potential function, its formal expression can be unavailable in several real-world applications. For this reason, we propose a two-layer Nash equilibrium seeking scheme in which a central coordinator exploits noisy feedback from the agents to design personalized incentives for them. By making use of those incentives, the agents compute a solution to an extended game, and then return feedback measures to the coordinator. We show that our algorithm returns an equilibrium if the coordinator is endowed with standard learning policies, and corroborate our results on a numerical instance of a hypomonotone game.

Keywords: Linear systems, Optimal control, Game theoretical methods
Abstract: Motivated by the fact that optimal performance criteria are often not known a priori, we present an approach to represent quadratic objective functions in the context of optimal control directly using finite, open-loop, non-optimal data trajectories of the state, input and a performance variable. Combined with a data-based representation of linear time-invariant systems this allows us to solve linear quadratic regulator problems with unknown dynamics and unknown cost matrices via data-dependent convex programmes. We show that this result is relevant to a specific class of linear quadratic games, in which one player is missing information regarding the control objectives of the other players and/or the system dynamics. The applicability of the presented results is highlighted via an example concerning human-robot interaction.

Keywords: Concensus control and estimation, Algebraic/geometric methods, Lyapunov methods
Abstract: In this work, the second-order Laplacian flow in Euclidean space which is a standard algorithm for consensus of double integrator systems is generalized to the abstract setting of Lie groups. For double integrator systems on a Lie group, by using gradients of Polar Morse functions whose critical points form a discrete subgroup, it is proved that consensus is achieved for almost all initial conditions of the agents whose connectivity is described by a nearest neighbor network. In this general framework, it turns out that the standard Euclidean second-order Laplacian flow and the Kuramoto oscillator are special cases in Euclidean space and the unit circle respectively.

Keywords: Concensus control and estimation, Distributed control, Machine learning
Abstract: This paper considers the consensus of mixed integer linear programming (MILP) problems as occurring in distributed control and machine learning of multi-agent networks. Unlike existing work on consensus problems, in which the agents only have to agree on the continuous part of their decision variables, this paper proposes a new method to enable them to also agree on the integer part. This mixed-integer setting may arise from distributed control problems of hybrid dynamical systems, or distributed machine learning problems using support vector machines. It is shown in this paper that the consensus of mixed-integer variables is guaranteed to be achieved by a tailored series of continuous consensus problems.

Keywords: Concensus control and estimation, Lyapunov methods, LMI's/BMI's/SOS's
Abstract: While reaching an agreement in multi-agent systems (MAS) can be ensured by enforcing some connectivity properties between agents, the consensus depends on their initial conditions and the network topology. In this context, our main objective in this paper is to sway the consensus value of multi-agent systems towards a desired value. The asymptotic stability and maximization of the domain of attraction for the bilinear model representing the opinion dynamics in the presence of limited control action for a fixed and connected network are investigated. By using algebraic graph theory and linear matrix inequality (LMI), we provide sufficient conditions guaranteeing the convergence of agents toward the desired consensus. Furthermore, examples illustrate the effectiveness of the proposed method.

Keywords: Distributed estimation over sensor nets, Concensus control and estimation, Uncertain systems
Abstract: Distributed estimation over sensor networks is one of the fundamental cooperative tasks involving multi-agent systems. Combining the Kalman filter with a consensus protocol is among the most successful strategies to address this problem. However, the availability of exact models is usually assumed. In practice, the models are often subject to parametric uncertainties. In this paper, we propose a robust distributed Kalman consensus filter. We consider that both the target system and sensing models have norm-bounded uncertainties in all parameter matrices. As a benchmark, we first introduce a centralized filter obtained from a robust regularized least-squares estimation problem. Then, we apply the hybrid consensus on measurements and information approach to derive a fully distributed version of this filter. We further establish steady-state stability conditions for both estimators. We also show that, for quadratically stable systems, the filters have bounded estimation error variance. Through an illustrative example, we assess the performance of the proposed estimators and provide comparisons with other robust distributed strategies.

Keywords: Network analysis and control, Concensus control and estimation, Agents networks
Abstract: Laplacian dynamics on signed graphs have a richer behavior than those on nonnegative graphs. In particular, their stability is not guaranteed a priori. Consequently, also the time-varying case must be treated with care. In particular, instabilities can occur also when switching in a family of systems each of which corresponds to a stable signed Laplacian. In the paper we obtain sufficient conditions for such a family of signed Laplacians to form a consensus set, i.e., to be stable and converging to consensus for any possible switching pattern. The conditions are that all signed Laplacian matrices are eventually exponentially positive (a Perron-Frobenius type of property) and normal.

Keywords: Network analysis and control, Concensus control and estimation, Control over networks
Abstract: Graph is an abstract representation commonly used to model networked systems and structure. In problems across various fields, including computer vision and pattern recognition, and neuroscience, graphs are often brought into comparison (a process is called textit{graph matching}) or checked for symmetry. Friendliness property of the associated adjacency matrices, specified by their spectral properties, is important in deriving a convex relaxation of the (intractable) discrete graph matching problem. In this work, we study unfriendliness properties of symmetric graphs by studying its relation to the underlying graph structure. It is revealed that a symmetric graph has two or more subgraphs of the same topology, and are adjacent to the same set of vertices. We then show that if adjacency matrices of symmetric graphs have distinct eigenvalues then there exist eigenvectors orthogonal to the vector of all ones, making them unfriendly. Relation of graph symmetry to uncontrollability of multi-agent systems under agreement dynamics with one controlled node is revisited. Examples of both synthetic and real-world graphs are also given for illustrations.

Keywords: Cooperative control, Agents and autonomous systems, Autonomous systems
Abstract: This paper proposes a consensus protocol for multiple double integrators using binary relative position measurements under a detailed-balanced digraph. In this protocol, an agent only requires two values to measure the positive and negative relative position of its neighbors. Compared to the existing binary measurement-based protocols, the presented protocol is free from agents' velocity measurements. The non-smooth Lyapunov analysis is utilized to establish that consensus is guaranteed under the proposed protocol for any detailed-balanced digraph with any positive gains. With numerical simulations, the proposed protocol is compared with the previous binary measurement-based protocol to show that the performance under the proposed protocol is least affected despite the lack of velocity measurements.

Keywords: Machine learning, Filtering, Intelligent systems
Abstract: When signals are measured through physical sensors, they are perturbed by noise. To reduce noise, low-pass filters are commonly employed in order to attenuate high frequency components in the incoming signal, regardless if they come from noise or the actual signal. Therefore, low-pass filters must be carefully tuned in order to avoid significant deterioration of the signal. This tuning requires prior knowledge about the signal, which is often not available in real-world applications. In order to overcome this limitation, we propose an adaptive low-pass filter based on Gaussian process regression. By considering a constant window of previous observations, updates and predictions fast enough for real-world filtering applications can be realized. Moreover, the online optimization of hyperparameters leads to an adaptation of the low-pass behavior, such that no prior tuning is necessary. We show that the estimation error of the proposed method is uniformly bounded, and demonstrate the flexibility and efficiency of the approach in several simulations.

Keywords: Uncertain systems, Machine learning, Sliding mode control
Abstract: Presence of model uncertainties creates challenges for model-based control design, and complexity of the control design is further exacerbated when coping with nonlinear systems. This paper presents a sliding mode control (SMC) design approach for nonlinear systems with partially known dynamics by blending data-driven and model-based approaches. First, an SMC is designed for the available (nominal) model of the nonlinear system. The closed-loop state trajectory of the available model is used to build the desired trajectory for the partially known nonlinear system states. Next, a deep policy gradient method is used to cope with unknown parts of the system dynamics and adjust the sliding mode control output to achieve a desired state trajectory. The performance (and viability) of the proposed design approach is finally examined through numerical examples.

Keywords: Chemical process control, Machine learning, Neural networks
Abstract: A new online steady-state detection (SSD) method based on convolutional neural networks has been developed. It mimics how humans carry out SSD by visually investigating measurement plots. The method is tested on synthetic and real-life data. We also investigate how it interacts with steady-state real-time optimization and affects its performance on a digital twin. The proposed method has comparable performance to traditional SSD methods; however, it is more intuitive to tune.

Keywords: Predictive control for nonlinear systems, Machine learning, Robust control
Abstract: Machine learning methods, like Gaussian process regression, allow improving the performance of model-based control methods, such as model predictive control. They can, for example, be used to improve the prediction quality of the used model of the system, learning the uncertain system part. However, fusing model-based approaches with machine learning methods raises the question of stability and per- formance guarantees. This work considers Gaussian process regression to learn uncertain system parts for nonlinear Lur’e type systems. More precisely, the Gaussian process is used to learn the nonlinearity of the system. Based on the learned model, it is shown how to compute suitable terminal regions to guarantee closed-loop stability. To do so, reproducing kernel Hilbert space membership of the learned function is exploited to compute high-probability sector bounds for the nonlinearity. From the obtained bounds, a high-probability terminal region that guarantees closed-loop stability is derived by methods from robust control theory. A flexible joint robotic arm example underlines the effectiveness of the approach.

Keywords: Machine learning, Optimization, Adaptive control
Abstract: We apply Gaussian process regression to the problem of centrifugal compressor performance modelling using hyperparameter optimisation. We test the proposed approach for compressor pressure ratio and efficiency prediction using compressor maps from four different machines and compare the performance of Gaussian process regression with multivariate polynomial regression. Gaussian process regression is found to outperform multivariate polynomial regression for this task especially when a small number of training samples are available. The proposed approach can therefore be a suitable method for online adaptation of centrifugal compressor maps for control and optimisation applications when dealing with fouling and other sources of compressor performance degradation.

Keywords: Machine learning, Agents and autonomous systems, Decentralized control
Abstract: Trust facilitates collaboration and coordination in teams and is paramount to achieve optimality in the absence of direct communication and formal coordination devices. We investigate the influence of agents' risk-attitudes on trust and emergence of coordination in multi-agent environments. To that end, we consider Independent Risk-sensitive Policy Gradient, Risk-sensitive REINFORCE, RL-agents in repeated 2-agent coordination games (Stag-Hunt). We experimentally validate our hypothesis that the agents' risk-attitudes influence coordination and collaboration by influencing the agents' learning dynamics and can lead to efficient learning of Pareto optimal policies. This suggests that risk-sensitive agents could achieve better results in multi-agent task environments.

Keywords: Neural networks, Machine learning, Predictive control for linear systems
Abstract: Convex piecewise quadratic (PWQ) functions frequently appear in control and elsewhere. For instance, it is well-known that the optimal value function (OVF) as well as Q-functions for linear MPC are convex PWQ functions. Now, in learning-based control, these functions are often represented with the help of artificial neural networks (NN). In this context, a recurring question is how to choose the topology of the NN in terms of depth, width, and activations in order to enable efficient learning. An elegant answer to that question could be a topology that, in principle, allows to exactly describe the function to be learned. Such solutions are already available for related problems. In fact, suitable topologies are known for piecewise affine (PWA) functions that can, for example, reflect the optimal control law in linear MPC. Following this direction, we show in this paper that convex PWQ functions can be exactly described by max-out-NN with only one hidden layer and two neurons.

Keywords: Agents networks, Network analysis and control, Concensus control and estimation
Abstract: Bounded confidence opinion dynamics are dynamic networks in which agents are connected if their opinions are similar, and each agent updates her opinion as the average of the neighbors' opinions. In homogeneous asymmetric Heglselmann-Krause (HK) models, all agents have the same confidence thresholds which could be different for the selection of upper and lower neighbors. This paper provides conditions for the convergence of the opinions of this class of HK models to consensus and to clustering. A new tighter bound on the time interval for reaching the steady state is also provided.

Keywords: Game theoretical methods, Control over networks, Distributed control
Abstract: We consider a network of autonomous agents, each described by a class of equilibrium-independent passive dynamics. Each agent seeks to minimize a coupled cost function subject to coupled inequality constraints over the output of its dynamics in steady-state. We propose an interior point based method with time-varying penalty function in order to achieve exact convergence to the generalized Nash equilibrium of the game while simultaneously guaranteeing constraint satisfaction for all time. This method is developed in both the full- and partial-information settings. Applications to velocity synchronization are provided.

Keywords: Network analysis and control
Abstract: Motivated by the increasing number of COVID-19 cases that have been observed in many countries after the vaccination campaign and relaxation of non-pharmaceutical interventions (NPIs), we propose a network model for the spread of recurrent epidemic diseases in a partially vaccinated population. The model encapsulates several realistic features, such as different vaccine efficacy against transmission and development of severe symptoms, testing practices, implementation of NPIs, isolation of detected individuals, and human behaviour. Using a mean-field approach, we analytically derive the epidemic threshold of the model and, if the system is above such a threshold, we compute the epidemic prevalence at the endemic equilibrium. These theoretical results show that precautious human behaviour and effective testing practices are key toward avoiding epidemic outbreaks. Interestingly, we found that, in many realistic scenarios, vaccination is successful in mitigating the outbreak by reducing the prevalence of seriously ill patients, but it could be a double-edged sword, favouring resurgent outbreaks, and it thus calls for higher testing rates, more cautiousness and responsibility among the population, or the reintroduction of NPIs to achieve full eradication.

Keywords: Network analysis and control, Optimal control, Control over networks
Abstract: In this paper, we apply an optimal Linear Quadratic (LQ) controller, which has an inherent structure that allows for a distributed implementation, to an irrigation network. The network consists of a water reservoir and connected water canals. The goal is to keep the levels close to the set-points when farmers take out water. The LQ controller is designed using a first-order approximation of the canal dynamics, while the simulation model used for evaluation uses third-order canal dynamics. The performance is compared to a P controller and an LQ controller designed using the third-order canal dynamics. The structured controller outperforms the P controller and is close to the theoretical optimum given by the third-order LQ controller for disturbance rejection.

Keywords: Control over networks, Network analysis and control
Abstract: A generalization of stability criteria for packet-based networked systems with state feedback is introduced. It is based on the small gain theorem for the case with causal and acausal sub-systems. Special emphasis is put on the consideration of the effects of individual bounded but time-varying packet delays together with the used packet skipping and hold mechanisms at the receiver sides. A state filtered Smith predictor with a PI state controller and a packet-based communication network is utilized in a simulation example to point out the easy-to-check nature of the proposed criterion.

Keywords: Optimal control of communication networks, Control over networks, Control over communication
Abstract: This paper considers an Linear Quadratic Regulator (LQR) design problem for multi-agent distributed control systems where designing an optimal feedback controller by considering communications among agents is desired for the reduction of communication burden in a network. To aim this, we deal with an LQR minimization problem with a regularization for sparse feedback matrix, where the sparsity in the feedback matrix is related to the reduction of the communication links in the multi-agent distributed control systems. We propose a simple but efficient iterative algorithm, so-called Iterative Shrinkage-Thresholding Algorithm (ISTA) for sparse LQR optimal control design. The proposed method can provide a trade-off solution between LQR cost and sparsity level on feedback matrix. Through various numerical experiments, we demonstrate that our proposed method can outperform the previous work using the Alternating Direction Method of Multiplier (ADMM) in terms of computational speed. Additionally, based on our proposed method, we introduce its deep neural network model, which can further improve the performance of the proposed algorithm in convergence speed.

Keywords: Optimization algorithms, Distributed cooperative control over networks, Agents networks
Abstract: In this paper we propose a distributed optimization problem with a global objective given by a weighted sum of local objectives, where each local weight encodes the absolute relevance of the local objective associated to an agent. In our settings, each agent is assumed to have only local (possibly incosistent) information regarding the relative importance of its objective function with respect to its neighboring agents. Indeed,this allows to model scenarios where only partial knowledge is available to each agents, e.g., for privacy reasons. In this regard, we propose a distributed framework where agents cooperate to both negotiate their absolute relevance and solve the resulting optimization problem. The proposed framework ensures finite-time convergence under the assumption that for each local objective function the related Hessian matrix has eigenvalues that are lower-bounded by a known constant.