Critic-Identifier Structure-Based ADP for Decentralized Robust Optimal Control of Modular Robot Manipulators

Abstract

This paper presents a decentralized robust optimal control method for modular robot manipulators (MRMs) via a novel critic-identifier (CI) structure-based adaptive dynamic programming (ADP) scheme. The robust control problem of MRMs is transformed into an optimal compensation control approach, which combines model-based compensation control, identifier-based learning control and ADP-based optimal control. The dynamic model of MRMs is formulated based on a torque sensing technique that is deployed for each joint module, where the local dynamic information is utilized effectively to design the model compensation controller. A neural network (NN) identifier is established to approximate the dynamics of the interconnected dynamic coupling (IDC). Based on the ADP algorithm, the Hamiltonian-Jacobi-Bellman (HJB) equation can be solved by constructing a critic NN, and the approximate optimal control policy is derived. The closed-loop robotic system is guaranteed to be asymptotic stable by the implementation of a set of decentralized control policies that have been developed. Finally, simulations verify the effectiveness of the proposed method.