The research results described present the performance of the Generalized Predictive Control (GPC) algorithm with a changing estimator and predictor model order for a specific application. The application is a hydraulically actuated heavy duty manipulator. Hydraulically actuated robotic manipulators, used in the large resource based industries, have a complex dynamic response in which, primarily due to the hydraulic actuator subsystems, the order of the dynamic model is not initially known and can change as the manipulator is operated. A nonlinear simulation model of the manipulator system is utilized in the work and the GPC controller is implemented with a CARIMA estimator together with an on-line, gradient based estimator model order determination technique. The results given show that with proper use of the order determination technique cost function and tuning of the GPC parameters, good performance and stability can be achieved.

This paper reports the results of an experimental study, which was conducted to evaluate the performance and implementation aspects of a generalized predictive control (GPC) technique to an electro-hydraulic positioning actuator. Poor dynamics and high nonlinearities form part of the difficulty in the control of hydraulic functions which make the application of adaptive controls an attractive solution. The applicability of GPC to the position control of hydraulic manipulator has been investigated through computer simulations in the literature. However, there is no experimental record of applying this technique to an actual hydraulic system. A suitable plant model is established and recursive U-D factorization technique is adopted for on-line estimation of time-varying plant parameters. Experimental results are obtained from a laboratory electrohydraulic actuator test stand. Various benchmark tests, comprising step and tracking inputs, demonstrate good performance and the promise of the technique. In spite of significant actuator dynamics (control voltage saturation, flow nonlinearity and dry frictional nonlinearity in the hydraulic actuator), successful control tests could be performed repetitively.