EECS661 Modeling and Optimal Control of Hybrid Systems: Two Case Studies

Biju Edamana, Scott Moura
December 16, 2008

Intro to Discrete Event Systems (EECS 661)
Professor Stephane Lafortune
University of Michigan
This report investigates a modeling formalism for a class of discrete-time hybrid systems that is amenable to servo-level control design techniques, known as mixed logical dynamical systems (MLD). During this course we discussed hybrid automata as one potential modeling framework for hybrid systems. However, a key limitation of this modeling approach is that control design is generally not straightforward or computer-implementable, for either supervisory or servo-level control. As a result, we studied MLD systems for this project, using Bemporad and Morari as our principle reference. The key advantage of the MLD modeling format is that it is amenable to numerical optimization routines for control optimization problems. In this report we apply both the hybrid automata and MLD modeling paradigms to systems from our own individual research projects. Once the individual systems are in MLD form, we then apply optimal control using mixed integer quadratic programming (MIQP).