Note: This content is accessible to all versions of every browser. However, this browser does not seem to support current Web standards, preventing the display of our site's design details.


Online Solvers for Hybrid MPC Using Generalized Disjunctive Programming



Damian Frick, Alexander Domahidi

Model predictive control (MPC) based on hybrid models that contain both continuous dynamics and logical statements is a very powerful tool for obtaining high-performance controllers that can handle most systems in practice.
However, the solution of the hybrid MPC problem is NP-hard in general, and it remains unclear how to solve these problems in acceptable run-time on embedded systems. One way is to reformulate them as mixed-integer binary programs and apply standard methods such as branch-and-bound. The reformulation is intuitive but has one major drawback: its relaxations are not very tight in practice, which leads to worse run-time.

The goal of this thesis is to examine the paradigm of Generalized Disjunctive Programming, which allows one to compute tighter relaxations for the case of hybrid MPC problems. The questions to be investigated are: (1) how can we efficiently compute the convex hulls of the relaxations, (2) can we achieve significant speed-ups by using these tighter relaxations and (3) can the class of systems be restricted such that certain properties can be universally exploited?

This project has a strong research component, and the candidates are expected to be strong in math and programming. Courses from mathematical optimization are a prerequisite.

The embedded optimization group at IfA develops methods and tools for numerical optimization on embedded hardware, mostly for use in the model predictive control context.

Weitere Informationen

Manfred Morari

Art der Arbeit: Masterthesis
50% Theory
30% Implementation
10% Testing
10% Documentation
Voraussetzungen: Optimization
Anzahl StudentInnen: 1
Status: done
Projektstart: Spring 2016