Model Predictive Control

  Illustration of the principle od model predictive control Copyright: © IC

Online optimization-based control approaches based on receding horizon information, such as model predictive control (MPC), have received a lot of attention in both academia and industrial application. In MPC, a control action is computed by solving at each time instant a suitable finite horizon open-loop optimal control problem in a receding horizon fashion. Only the first input of the resulting optimal sequence is applied to the considered plant.

MPC schemes are typically based on the assumption that the optimal control input can be computed exactly and in negligible time at each sampling instant. However, in the presence of fast system dynamics or limited hardware resources, this assumption may no longer be valid. Establishing stability or performance guarantees becomes then a challenging task when it comes to the practical implementation of MPC schemes. This advocates for developing efficient MPC algorithms that allow for a rapid computation of the optimal control input, or at least of a sufficiently good approximation of it, while still being able to provide guarantees on system-theoretical closed-loop properties (e.g. stability and constraint satisfaction). Here, the closed-loop system consists of the combined dynamics of the plant and the optimization algorithm.

Our goal is to develop and analyze so-called anytime barrier function based MPC approaches in conjunction with suitable optimization algorithms. By taking the dynamics of the underlying optimization into account, the resulting MPC algorithms allow to guarantee properties like closed-loop stability and constraint satisfaction for arbitrary fast system dynamics. Recent implementations in autonomous driving tasks have shown that the developed approach performs very well in practice.


Related Publications

Title Authors and Contributors
Anytime MHE-based output feedback MPC
In: IFAC-PapersOnLine, 54 (2021), 6, 264-271
Contribution to a conference proceedings, Journal Article
[DOI: 10.1016/j.ifacol.2021.08.555]
Gharbi, Meriem (Corresponding author)
Ebenbauer, Christian Johannes (Corresponding author)
Sparsity-Exploiting Anytime Algorithms for Model Predictive Control: A Relaxed Barrier Approach
In: IEEE transactions on control systems technology, 28 (2018), 2, 425-435
Journal Article
[DOI: 10.1109/TCST.2018.2880142]
Feller, Christian (Corresponding author)
Ebenbauer, Christian Johannes
Online learning with stability guarantees: A memory-based warm starting for real-time MPC
In: Automatica : a journal of IFAC, 122 (2020), 109247
Journal Article
[DOI: 10.1016/j.automatica.2020.109247]
Schwenkel, Lukas (Corresponding author)
Gharbi, Meriem
Trimpe, Johann Sebastian
Ebenbauer, Christian Johannes
A stabilizing iteration scheme for model predictive control based on relaxed barrier functions
In: Automatica : a journal of IFAC, the International Federation of Automatic Control, 80 (2017), 328-339
Journal Article
[DOI: 10.1016/j.automatica.2017.02.001]
Feller, Christian (Corresponding author)
Ebenbauer, Christian Johannes (Corresponding author)