A Neural-network-based Nonlinear Controller for Robot Manipulators with Gain-learning Ability and Output Constraints

This paper presents an adaptive robust controller for tracking control problems of robotic manipulators with output constraints. To realize the constrained objective, the controller is designed using a modified backstepping scheme. A neural-network model is employed to deal with uncertain nonlinearities and disturbances inside the system dynamics. To accomplish an asymptotically tracking performance under predefined constraints, both the control gains and the neural-network approximator are self-learnt with nonlinear laws. Effectiveness of the proposed controller is carefully verified by theoretical proofs and comparative simulation results.