Experimentally validated repetitive-predictive control of a robot arm with constraints

A multivariable predictive-repetitive controller has recently been reported that uses a frequency response decomposition technique to detect the dominant frequency components in the reference signal for each output and embeds these in the controller. This paper addresses issues concerning the implementation of operational constraints on control signal amplitudes, their increments and on the plant outputs. The major contribution is an on-line solution developed using quadratic programming via identification of active constraints. Supporting experimental results from application to a 2D robotic system are also given.