This work studies a multi-player H∞ differential game for systems of general linear dynamics. In this game, multiple players design their control inputs to minimize their cost functions in the presence of worst-case disturbances. We first derive the optimal control and disturbance policies using the solutions to Hamilton-Jacobi-Isaacs (HJI) equations. We then prove that the derived optimal policies stabilize the system and constitute a Nash equilibrium solution. Two integral reinforcement learning (IRL) -based algorithms, including the policy iteration IRL and off-policy IRL, are developed to solve the differential game online. We show that the off-policy IRL can solve the multi-player H∞ differential game online without using any system dynamics information. Simulation studies are conducted to validate the theoretical analysis and demonstrate the effectiveness of the developed learning algorithms.

This book offers a thorough introduction to the basics and scientific and technological innovations involved in the modern study of reinforcement-learning-based feedback control. The authors address a wide variety of systems including work on nonlinear, networked, multi-agent and multi-player systems. A concise description of classical reinforcement learning (RL), the basics of optimal control with dynamic programming and network control architectures, and a brief introduction to typical algorithms build the foundation for the remainder of the book. Extensive research on data-driven robust control for nonlinear systems with unknown dynamics and multi-player systems follows. Data-driven optimal control of networked single- and multi-player systems leads readers into the development of novel RL algorithms with increased learning efficiency. The book concludes with a treatment of how these RL algorithms can achieve optimal synchronization policies for multi-agent systems with unknown model parameters and how game RL can solve problems of optimal operation in various process industries. Illustrative numerical examples and complex process control applications emphasize the realistic usefulness of the algorithms discussed. The combination of practical algorithms, theoretical analysis and comprehensive examples presented in Reinforcement Learning will interest researchers and practitioners studying or using optimal and adaptive control, machine learning, artificial intelligence, and operations research, whether advancing the theory or applying it in mineral-process, chemical-process, power-supply or other industries.

This handbook presents state-of-the-art research in reinforcement learning, focusing on its applications in the control and game theory of dynamic systems and future directions for related research and technology. The contributions gathered in this book deal with challenges faced when using learning and adaptation methods to solve academic and industrial problems, such as optimization in dynamic environments with single and multiple agents, convergence and performance analysis, and online implementation. They explore means by which these difficulties can be solved, and cover a wide range of related topics including: deep learning; artificial intelligence; applications of game theory; mixed modality learning; and multi-agent reinforcement learning. Practicing engineers and scholars in the field of machine learning, game theory, and autonomous control will find the Handbook of Reinforcement Learning and Control to be thought-provoking, instructive and informative.

Neural Networks for Control brings together examples of all the most important paradigms for the application of neural networks to robotics and control. Primarily concerned with engineering problems and approaches to their solution through neurocomputing systems, the book is divided into three sections: general principles, motion control, and applications domains (with evaluations of the possible applications by experts in the applications areas.) Special emphasis is placed on designs based on optimization or reinforcement, which will become increasingly important as researchers address more complex engineering challenges or real biological-control problems.A Bradford Book. Neural Network Modeling and Connectionism series

The significantly expanded and updated new edition of a widely used text on reinforcement learning, one of the most active research areas in artificial intelligence. Reinforcement learning, one of the most active research areas in artificial intelligence, is a computational approach to learning whereby an agent tries to maximize the total amount of reward it receives while interacting with a complex, uncertain environment. In Reinforcement Learning, Richard Sutton and Andrew Barto provide a clear and simple account of the field's key ideas and algorithms. This second edition has been significantly expanded and updated, presenting new topics and updating coverage of other topics. Like the first edition, this second edition focuses on core online learning algorithms, with the more mathematical material set off in shaded boxes. Part I covers as much of reinforcement learning as possible without going beyond the tabular case for which exact solutions can be found. Many algorithms presented in this part are new to the second edition, including UCB, Expected Sarsa, and Double Learning. Part II extends these ideas to function approximation, with new sections on such topics as artificial neural networks and the Fourier basis, and offers expanded treatment of off-policy learning and policy-gradient methods. Part III has new chapters on reinforcement learning's relationships to psychology and neuroscience, as well as an updated case-studies chapter including AlphaGo and AlphaGo Zero, Atari game playing, and IBM Watson's wagering strategy. The final chapter discusses the future societal impacts of reinforcement learning.

This book presents a class of novel optimal control methods and games schemes based on adaptive dynamic programming techniques. For systems with one control input, the ADP-based optimal control is designed for different objectives, while for systems with multi-players, the optimal control inputs are proposed based on games. In order to verify the effectiveness of the proposed methods, the book analyzes the properties of the adaptive dynamic programming methods, including convergence of the iterative value functions and the stability of the system under the iterative control laws. Further, to substantiate the mathematical analysis, it presents various application examples, which provide reference to real-world practices.