Robust model‐free adaptive iterative learning formation for unknown heterogeneous non‐linear multi‐agent systems

Abstract

The formation problem for a class of non-linear multi-agent systems (MASs) is considered under bounded disturbance and repetitive environment. The agent dynamics here is unknown and heterogeneous, and the disturbances are both time-varying and iteration-varying. Based on a novel disturbance-related dynamic linearisation technique, the agent dynamics is first transformed into an equivalent dynamic linearisation data model along the iteration axis. Then, distributed robust model-free adaptive iterative learning protocols are designed to drive the agents to the desired formation pattern. By theoretic analysis, the formation tracking error of the MASs is ultimately bounded and the result is further extended to an iteration-varying topology and pattern case. As a data-driven method, the protocol design has the feature that only the input and output data of neighbouring agents are used. Finally, numerical examples are presented to show the effectiveness and robustness of this method.