Nonlinear model predictive iterative learning control for robotic system

A nonlinear model predictive controller based on iterative learning control (NMPILC) is proposed. The nonlinear plant dynamic is described by a fuzzy model which contains local liner models. Based on this model, model predictive control algorithm that utilizes past data along with real-time measurements is devised. The proposed control scheme takes advantages of the iterative learning law and model predictive control, which consists of time direction information and an iterative learning term. This algorithm is developed to address the learning rate for a class of repetitive system with non-repetitive disturbances. The iterative learning control law is given. Simulation on a single-joint mechanical arm shows the effectiveness of the proposed NMPILC. Compared with the exiting model predictive iterative learning control (MPILC), the results obtained in the experiment have quicker convergence rate.