Distributed data-driven iterative learning consensus tracking for unknown multi-agent systems using evolutionary neural networks

This paper provides a data-driven distributed parameter adaptive iterative learning consensus tracking strategy for nonlinear nonaffine discrete-time multi-agent systems with unknown dynamics. By transforming the ideal learning controller on the timeline into a direct iterative learning control strategy in the iterative domain, the design of the control protocol is only data-driven. Unlike existing parameter tuning control methods, the parameter tuning approach presented in this paper adjusts the parameters online through topological information, eliminating the need for multiple trials and adjustments based on experience. The gain time variability of multi-agent systems is learned and compensated by the extended generalized regression neural networks evolution control. By introducing a limited incremental evolution mechanism, the optimal control parameters can be adjusted online during the control process to find the system trajectory to achieve optimal output synchronization, so as to improve the control efficiency of iterative learning control. Different from the existing directed fixed topology works of multi-agent systems, the consensus convergence properties of fixed directed communication topologies and iterative time-varying communication topologies along the iterative domain are established by contraction mapping theorem. Two numerical simulation examples are conducted to validate the effectiveness of the proposed control protocol.

Authors

• Kechao Xu
• Bo Meng
• Zhen Wang

Keywords

• Data-driven
• Multi-agent systems
• Evolutionary neural networks
• Generalized regression neural networks
• Iterative learning control