Critic-Identifier Structure ADP Based Near-optimal Decentralized Tracking Control of Modular and Reconfigurable Robots

This paper investigates a near-optimal decentralized tracking control (DTC) scheme for modular and reconfigurable robots (MRRs) via a novel critic-identifier (CI) structure-based adaptive dynamic programming (ADP) algorithm. The DTC problem of MRRs is transformed into an optimal control issue, which consists of local namely controller, local optimal feedback controller and robust compensator. By using desired states of coupled subsystems to substitute their corresponding actual states, the strict norm-boundedness assumption of interconnections can be avoided. By using self-learning ability of neural network (NN), an identifier is constructed to approximate the subsystem dynamics. Then, the local desired control law can be obtained based on identified dynamics. A critic NN is constructed to solve Hamiltonian-Jacobi-Bellman (HJB) equation, and the local tracking feedback control policy is derived. To remove the overall error caused by the substitution, identification and approximation of critic NN, a robust term is added to guarantee the reliable performance of MRR system. The stability of the closed-loop system is ensured to be asymptotically stable by using the Lyapunov's direct methods. Finally, simulation studies show the effectiveness of the proposed scheme.