Two-Level Game-Based Distributed Optimal Fault-Tolerant Control for Nonlinear Interconnected Systems
Yuhang Xu,Bin Jiang,Hao Yang
DOI: https://doi.org/10.1109/tnnls.2019.2958948
IF: 14.255
2020-11-01
IEEE Transactions on Neural Networks and Learning Systems
Abstract:This article addresses the distributed optimal fault-tolerant control (FTC) issue by using the two-level game approach for a class of nonlinear interconnected systems, in which each subsystem couples with its neighbors through not only the states but also the inputs. At the first level, the FTC problem for each subsystem is formulated as a zero-sum differential game, in which the controller and the fault are regarded as two players with opposite interests. At the second level, the whole interconnected system is formulated as a graphical game, in which each subsystem is a player to achieve the global Nash equilibrium for the overall system. The rigorous proof of the stability of the interconnected system is given by means of the cyclic-small-gain theorem, and the relationship between the local optimality and the global optimality is analyzed. Moreover, based on the adaptive dynamic programming (ADP) technology, a distributed optimal FTC learning scheme is proposed, in which a group of critic neural networks (NNs) are established to approximate the cost functions. Finally, an example is taken to illustrate the efficiency and applicability of the obtained theoretical results.
computer science, artificial intelligence, theory & methods,engineering, electrical & electronic, hardware & architecture