Hierarchical Optimization Prescribed Performance Framework for Networked Mobile Manipulators: A Novel Reinforcement Learning Approach

This paper investigates distributed optimal teleoperation control for networked mobile manipulators (NMMs) subject to model uncertainties, nonholonomic constraints, and external disturbances. To achieve cost-minimization cooperative control with prescribed transient and steady-state performance, a hierarchical optimization prescribed performance (HOPP) framework is proposed by integrating a reinforcement learning-based optimization estimator (RLOE) with a prescribed performance stability controller (PPSC). In the proposed scheme, the RLOE generates distributed reference trajectories for slave mobile manipulators through local neighbor interactions while minimizing a cooperative performance index. The PPSC is then designed to guarantee bounded tracking errors with prescribed convergence behavior for both the master and slave manipulators. Lyapunov-based analysis is provided to establish the boundedness and convergence properties of the closed-loop system. Rooted in Lyapunov stability theory, the proposed control algorithm is designed to ensure reliability and efficacy. Simulation studies on a teleoperation system of 2-DoF mobile manipulators demonstrate that the proposed method achieves accurate tracking, reduced cooperative cost, and improved transient performance.