Learning-Based Data-Driven Cooperative Adaptive Fault-Tolerant Tracking of Heterogeneous Multiagent Systems

In this paper, the problem of cooperative output tracking for heterogeneous multi-agent systems with unknown model parameters and actuator faults is investigated. First, based on the output regulation theory and Lyapunov method, a model-based distributed fault-tolerant control framework is designed to ensure that the tracking errors can still converge asymptotically under actuator faults. Then, to remove the constraints of system model parameters on the framework, a data-driven online learning technique is developed to learn the parameters of fault-tolerant controllers and the solution to the output regulator equations by using the collected systems data. Moreover, two termination thresholds are constructed for online learning based on the properties of the designed Riccati equation. Theoretical analysis shows that the parameters and the solution to the Riccati equation obtained by satisfying the thresholds ensure the stability of the distributed fault-tolerant control framework even if there is a small error. Finally, the data-driven model-free cooperative fault-tolerant control algorithm is obtained by combining the above framework and online learning, and the effectiveness and advantages of the algorithm are verified by an example of multi-autonomous vehicles.