Decentralized robust zero-sum neuro-optimal control for modular robot manipulators in contact with uncertain environments: theory and experimental verification

Abstract

This paper presents a decentralized robust zero-sum optimal control approach for modular robot manipulators (MRMs) in contact with uncertain environments based on the adaptive dynamic programming (ADP) algorithm. The dynamic model of MRMs is formulated via joint torque feedback technique that is deployed for each joint module to design the model compensation controller. An uncertainty decomposition-based robust control is developed to compensate the model uncertainties, and then, the robust optimal control problem of the MRM system is transformed into a two-player zero-sum optimal control one. According to the ADP algorithm, the Hamilton–Jacobi–Isaacs equation can be solved by establishing action and critic neural networks, thus making the derivation of the approximate optimal control policy feasible. Based on the Lyapunov theory, the closed-loop robotic system is proved to be asymptotic stable under the developed decentralized control method. Finally, experiments are conducted to verify the effectiveness and advantages of the proposed method.