A learning controller based on a nonlinear ARX inverse model identified by a genetic algorithm

摘要

Abstract A nonlinear model based feedforward and feedback controller is described. This controller can control a nonlinear plant, such as a robot, whose dynamics are unknown. In the feedforward controller, it employs a Genetic Algorithm to identify nonlinear difference equations that model the inverse dynamics of the plant from the measured forcing functions and responses. In the feedback part, a PD feedback controller is employed to compensate for unmodelled dynamics and unexpected disturbances. Furthermore, an add-on learning controller is established to reduce tracking errors for repetitive tasks. The controller is validated with the control of a simulated two-joint manipulator. Simulation results demonstrate that a NARX model based feedforward/feedback and learning controller provides very accurate tracking.