Multi-agent formation control method based on actor-reviewer reinforcement learning and fuzzy logic

Abstract

The invention provides a multi-agent formation control method based on actor-reviewer reinforcement learning and fuzzy logic. An optimal control method is introduced into a pilot follower formation control method of a multi-robot system, and the capability of approaching a continuous function of a fuzzy logic system is utilized to solve the problem that it is difficult to obtain an analytical solution by a Hamilton-Jacobian-Bellman equation in optimal control; meanwhile, an actor-reviewer reinforcement learning algorithm is combined to form an actor fuzzy logic system module and a reviewer fuzzy logic system module, the former executes control behaviors, and the latter evaluates the behaviors selected by the former and feeds evaluation information back to the former. According to the method, the control performance and the resource loss can be balanced, and the adaptability of the multi-robot system to the environment is improved in an online learning mode.

Description

technical field [0001] The invention belongs to the field of robot formation control, and in particular relates to a multi-agent second-order linear system optimal formation control method based on actor-critic reinforcement learning and fuzzy logic. Background technique [0002] In the past ten years, due to its greater redundancy, multi-robot systems have better fault tolerance and robustness than single-robot systems, and can cooperate to complete many tasks that single robots cannot complete. In a multi-robot system, robot formation is one of the control methods for robots to perform tasks cooperatively. As one of the formation control techniques, the leader-follower method can realize the distributed control of multi-robot systems, and has the characteristics of high flexibility and easy use. Introducing the optimal control method into the multi-robot system formation control can achieve the control goal of balancing control performance and resource consumption by mini...

AI Technical Summary

Problems solved by technology

It mainly takes the second-order linear system as the research object. Firstly, the optimal control method is introduced into the leader-follower formation control method of the multi-robot system, and the ability of the fuzzy logic system to approximate the continuous function is used to solve the problem of Hamilton-Jacobi-Bell in optimal control. It is difficult to find an analytical solution to the Mann equation; secondly, combined with the actor-critic reinforcement learning algorithm, an actor fuzzy logic system module and a critic fuzzy logic system module are formed, the former executes the control behavior, and the latter evaluates the behavior selected by the former And feedback the evaluation information to the former; Finally, the Bellman residual error is minimized by the gradient descent method, and the parameter vector update law of the critic fuzzy logic system module and actor fuzzy logic system module is designed

Images