Implementation Aspects of Model Predictive Control for Embedded Systems
In this paper we discuss implementation related aspects of model predictive control schemes on embedded platforms. Exemparily we focus on fast gradient methods and present results from an implementation on an embedded lowcost ARM processor. We show that input quantization taking place in actuators should be taken into account in order to determine the maximum number of iterations of the online optimization. Furthermore, we present results which allow to determine the online memory demand of the fast-gradient MPC algorithm on the embedded system offline. As a case study we consider a Segway-like robot, modeled by an LTI-system with 8 states and 2 inputs subject to box input constraints. The test system runs at a sampling rate of 4ms and uses MPC horizons up to 20 steps in a hard realtime system with limited CPU time and memory.