Header logo is

Learning an Approximate Model Predictive Controller with Guarantees




A supervised learning framework is proposed to approximate a model predictive controller (MPC) with reduced computational complexity and guarantees on stability and constraint satisfaction. The framework can be used for a wide class of nonlinear systems. Any standard supervised learning technique (e.g. neural networks) can be employed to approximate the MPC from samples. In order to obtain closed-loop guarantees for the learned MPC, a robust MPC design is combined with statistical learning bounds. The MPC design ensures robustness to inaccurate inputs within given bounds, and Hoeffding’s Inequality is used to validate that the learned MPC satisfies these bounds with high confidence. The result is a closed-loop statistical guarantee on stability and constraint satisfaction for the learned MPC. The proposed learning-based MPC framework is illustrated on a nonlinear benchmark problem, for which we learn a neural network controller with guarantees.

Author(s): Michael Hertneck and Johannes Koehler and Sebastian Trimpe and Frank Allgöwer
Journal: IEEE Control Systems Letters
Volume: 2
Number (issue): 3
Pages: 543-548
Year: 2018
Month: July

Department(s): Intelligent Control Systems
Bibtex Type: Article (article)
Paper Type: Journal

DOI: 10.1109/LCSYS.2018.2843682
State: Published

Links: arXiv
Attachments: PDF


  title = {Learning an Approximate Model Predictive Controller with Guarantees},
  author = {Hertneck, Michael and Koehler, Johannes and Trimpe, Sebastian and Allg{\"o}wer, Frank},
  journal = {IEEE Control Systems Letters},
  volume = {2},
  number = {3},
  pages = {543-548},
  month = jul,
  year = {2018},
  month_numeric = {7}