Neural Learning and Control Co‐Design for Robot Arm Control

Redundant manipulators have received intensive research focus for dexterous manipulation of complicated tasks. A redundant manipulator provides a nonlinear mapping from its joint space to the Cartesian workspace. This chapter presents a model‐free dual neural network for redundant manipulator control. Different from the pure learning problem of recurrent neural networks, the problem investigated in the chapter involves both learning and control, where they interact with each other. The chapter designs extra neural dynamics to learn the model of the manipulator, and uses the online learned results to achieve the model‐free control of manipulators. It presents the stability theorem and its theoretical proof. The chapter shows typical simulations on using the proposed neural network for the tracking of a circular motion. The simulations are conducted under the condition that no prior knowledge about the manipulator parameters is employed.