Neuro-based adaptive internal model control for robot manipulators

Abstract

The application of internal model control (IMC) for process control has received much attention during the past decade. In this paper, the application of IMC for robot control is investigated. Although the IMC approach is shown to be more robust compared to conventional robot control approaches, such as the computed-torque approach, its performance degrades in the presence of large modelling uncertainties and external disturbances. In this paper, a neuro-based adaptive internal model control scheme is proposed. Within the framework of this control structure, a backpropagation neural network algorithm is incorporated into a fired structure internal model controller for robot control. Simulation results, based on a two-link robot confirm the effectiveness of the proposed control algorithm even in the presence of large modelling uncertainties and external disturbances.