IJCAS

This paper develops a data-driven multiplayer mixed-zero-sum game control approach of modular robot manipulators (MRMs) with uncertain disturbance via adaptive dynamic programming (ADP). The dynamic model of MRMs is formulated via joint torque feedback. We deem n modules and uncertain disturbance as players in the game theory structure. The uncertainties in the MRM system such as joint friction and interconnected dynamic couplings (IDCs), we employ data-driven model based on recurrent neural networks (RNNs) to built. According to ADP, the Hamilton-Jacobi (HJ) equation can be solved by using critic neural networks and derivate the optimal control law. The closed-loop robotic system is proved to be asymptotic stable via mixed-zero-sum game control. Experiments are conducted to verify the effectiveness.

Keywords: Adaptive dynamic programming, mixed-zero-sum, modular robot manipulators, neural networks.