MOVES Seminar, 13 Jan 2010, 11:00

 

Ulrich Loup

 

Actuating Symbolically - A Case Study

 

Abstract:

 

This talk is concerned with a case study which was performed in cooperation with the Institut für Regelungstechnik of RWTH Aachen. We tackled a problem emerging in control theory: given a model of a real-world system to be controlled, find a function which always computes the optimal control for the system w.r.t. a specific prediction horizon. Such so-called control functions can be computed for linear models analytically, but this is in general not the case for non-linear models. In current applications such systems are controlled using dynamic programming in an approximate, time-discrete setting, particularly without computing control functions themselves. We tried to fill this gap by computing the optimal control using dynamic programming on a symbolic working point of the system. This approach yields a term representing the optimal control depending on an arbitrary system state -- a control function. The talk summarizes some experimental results and pinpoints problems regarding the symbolic computations.