Optimal Fault-Tolerant Containment Control of Multi-Agent Systems: An Adaptive Learning Approach

In this paper, we seek to address the optimal fault-tolerant containment control problem of multi-agent systems through a new adaptive learning approach. The fault-tolerant containment control problem is first transformed into a multi-agent nonzero-sum game, based on which the coupled Hamilton-Jacobi-Bellman (HJB) equation including an actuator efficiency factor and non-quadratic term is derived. It is proved that the solution to this coupled HJB equation satisfies the Nash equilibrium condition, and an adaptive learning approach based on actor-critic neural networks and adaptive projection is proposed to obtain the Nash equilibrium solution without knowing system models. Finally, a simulation example is presented to testify the feasibility of the proposed method.