A Non-Zero-Sum-Based Neural-Optimal control method for Modular and Reconfigurable Robot Systems

Abstract

Improving control performance of modular and reconfigurable robot(MRR) system while reducing the energy cost required by the controller, a non-zero-sum neural-optimal control algorithm is proposed. Control law of each joint module is designed as a participant, so that each participant achieves the optimal overall energy consumption under the game theory. optimal control law of the system under non-zero-sum game is obtained through the policy iteration method. Lyapunov method is used to prove stability. Numerical simulations proves superiority of the controller.