Model-free Adaptive Iterative Learning Control For Multi-agent Systems

For nonlinear discrete-time multi-agent systems(MAS),this paper focuses on the collaborative control of MAS with unknown models,and focuses on the bipartite containment control and bipartite consensus control of MAS.The main work and innovation points of the paper are summarized as follows:(1)Model free adaptive iterative learning bipartite containment control is researched for a class of MAS under the interaction of cooperation and antagonistic interaction.Firstly,model-unknown nonlinear systems become linear systems by using dynamic linearization methods;secondly,only using the input and output data of MAS to design the controller,this design avoids the trouble of making it difficult to build a suitable model for the system;then,we gives the conditions that can ensure that MAS realizes bipartite containment,so that the containment error becomes smaller and smaller with the increase of the number of iterations.Finally,the simulation verifies the correctness of the proposed controller.(2)The quantized model-free adaptive iterative learning control algorithm is proposed for solving the bipartite containment tracking problem of nonlinear multiagent systems with quantized communication.First,we use dynamic linearization method to transform a nonlinear multi-agent system into a linear data model.Secondly,the design of the quantization controller only requires the input and output data after quantization of the MAS,and does not depend on the system model.Finally,we give the conditions for achieving bipartite containment in MAS.The simulation results show that the output trajectory of the follower converges in the convex hull formed by the leader’s trajectory and its symmetric trajectory.(3)For solving the problem of bipartite consensus tracking of nonlinear MAS,an event-driven model-free adaptive iterative learning control method is proposed,in which the interaction between multiple agents includes collaboration and antagonistic.First,we use dynamic linearization to transform a nonlinear multi-agent system into a linear data model.Second,we design the controller without relying on the model information of the system.Finally,the simulation results show that the output state of the follower converges to the leader’s trajectory and its symmetric trajectory.