Efficient Implementation of Step Response Prediction Models for Embedded Model Predictive Control

Abstract

Different formulations of the step response prediction model used in Model Predictive Control (MPC) are examined in this paper. The advantages and disadvantages of the models are uncovered, and a class of model errors that lead to the deterioration of the prediction capability of step response models is identified. Particular attention is paid to the effect of small truncation errors, and the results show that such errors might lead to poor predictions, if the last element in the step response sequence is used to extend the prediction model. Several implementation aspects that are crucial for embedded targets with limited resources are also discussed. The results of the performance and computational enhancements proposed in this paper are analyzed in a simulation study.