Neural network-based adaptive controller design for robotic manipulator subject to varying loads and unknown dead-zone

In this article, aiming at handling the trajectory tracking issue of industrial manipulator system (IMS) with modeling uncertainty, varying loads (VL) and unknown dead-zone characteristic, a compensation-based adaptive switching controller synthesis is proposed. In this scheme, the dynamic model of the IMS under VL is regarded as a switched system (SS) with a specified modal set. The nonlinear term related to plant model in each subsystem is approximated by radial basis function neural network (RBFNN) so as to avoid the reliance of the controller on the accurate model, and the unknown dead-zone is estimated and compensated by NN, from which the corresponding NN robust compensation term is developed to eliminate the potential perturbations and estimated errors. The designed controller with switching mechanism effectively solves the problem of degradation of the tracking accuracy caused by VL. Finally, the uniform ultimate boundedness of error signals is analyzed by the average dwell time (ADT) approach, multi-Lyapunov function method and the synthesized adaptive control law, and the effectiveness of the developed scheme is verified by simulation.