Leader-Follower Formation Control For Indoor Wheeled Robots Via Dual Heuristic Programming

Abstract

As a class of reinforcement learning (RL) and approximate dynamics programming (ADP), dual heuristic programming (DHP), which relies on an actor-critic structure for approximating optimal costate and control policy, has been widely used for solving nonlinear optimal control problems. In this paper, we extend the DHP algorithm to formation control applications of indoor wheeled robots, where multiple nonlinear optimal control problems require to be solved and the interactions between multi-robots need to be concerned. Specifically, we propose a DHP-based formation control scheme (DHP-FC) in a leader-follower manner for indoor wheeled robots. In the proposed DHP-FC scheme, a nonlinear tracking problem of the leading robot is solved using DHP for following the trajectory generated by a virtual robot. Also, decentralized DHP-based following controllers are designed for realizing formation movement based on the real-time information received from the leader. Simulation results illustrate the potentiality of the DHPFC scheme under different reference trajectories and formation patterns.