Secure Distributed Estimation-Based Data-Driven Leader-Following Control for Discrete-Time MASs Under Deception Attacks and Sensor Faults

In this paper, we consider the leader-following consensus problem for a class of heterogeneous nonlinear discrete-time multi-agent systems (MASs) with unknown dynamics under network deception attacks and sensor faults. The deception attacks considered occur in the communication network between the leader and the followers, compromising the authenticity of the leader's information received by the followers and affecting the cooperative performance. Since not all followers can directly access the leader's information and their measurements may be influenced by deception attacks and sensor faults, a secure distributed estimation algorithm is proposed to allow each follower to obtain accurate information about the leader. To achieve indirect estimation of unknown fault signals, nonlinear autoregressive with exogenous input neural network (NARXNN)-based fault estimators are proposed, combining the open-loop and closed-loop modes of NARXNN. Moreover, considering that followers have unknown dynamics and exhibit heterogeneity, decentralized model-free adaptive controllers are designed based on the estimation of leader information and fault signals. Unlike methods that use consensus error for indirect consensus control, the proposed approach allows each agent to directly acquire a leader estimate and achieve direct leader-following consensus control using only local input/output data. Finally, simulation examples confirm the effectiveness of the proposed data-driven leader-following consensus control method.