Real-Time Optimization for Control of Mechatronic Systems

Moritz Diehl, Optimization in Engineering Center OPTEC, and ESAT, K.U. Leuven, Belgium

Abstract:

Model Predictive Control (MPC) is a powerful feedback control technique that is based on the fast online solution of
optimal control problems in real-time. With dedicated algorithms, solution times can be brought down to the millisecond
timescale, so that MPC is becoming feasible for applications in mechatronics, like control of combustion engines, machine
tools, or robot arms. The MPC problems can particularly fast and reliably be solved if they are convex, and special
emphasis must be put into a careful  problem formulation. In this talk, we outline recent progress on both the formulation
of convex MPC problems as well as the fast and reliable solution of MPC problems. We particularly discuss the problem of
time optimal robot control and show results obtained at a robot arm at KU Leuven.

(joint work with Diederik Verscheure, Hans Joachim Ferreau, Lieboud Vandenbrouck, Bram Demeulenaere, Hans Georg Bock, Jan
Swevers)