Project of U. Halldorsson, Ruhr Universitaet Bochum, Germany. 2000-2004.
A new approach is proposed to reduce the computational load of nonlinear model based predictive controllers. The idea is based on dividing a long prediction horizon into only a few equidistant intervals with piecewise constant control signals. After solving a first dynamic optimization problem the prediction horizon is divided into two parts, keeping the second part of the solution fixed and doubling the sampling rate in the first half of the control horizon. Using these settings a second optimization is performed to improve the first acquired solution. This procedure is repeated until the applied control step has a reasonable sampling time.