TR2021-126

Learning Disagreement Regions with Deep Neural Networks to Reduce Practical Complexity of Mixed-Integer MPC


    •  Chakrabarty, A., Quirynen, R., Romeres, D., Di Cairano, S., "Learning Disagreement Regions with Deep Neural Networks to Reduce Practical Complexity of Mixed-Integer MPC", IEEE International Conference on Systems, Man, and Cybernetics, October 2021.
      BibTeX TR2021-126 PDF Video
      • @inproceedings{Chakrabarty2021oct,
      • author = {Chakrabarty, Ankush and Quirynen, Rien and Romeres, Diego and Di Cairano, Stefano},
      • title = {Learning Disagreement Regions with Deep Neural Networks to Reduce Practical Complexity of Mixed-Integer MPC},
      • booktitle = {IEEE International Conference on Systems, Man, and Cybernetics},
      • year = 2021,
      • month = oct,
      • url = {https://www.merl.com/publications/TR2021-126}
      • }
  • MERL Contacts:
  • Research Areas:

    Control, Machine Learning, Optimization

Abstract:

Efficiently computing solutions to mixed-integer optimization-based control problems, such as in model predictive control (MPC) of hybrid systems, is extremely challenging due to the exponential worst-case complexity. The practical time-complexity of computing good control actions can be reduced by using a combination of two solvers: a strong solver that generates optimal or near-optimal closed-loop solutions with a large number of iterations, and a weak solver that converges quickly to suboptimal closed-loop solutions. In this paper, we propose the use of deep neural networks to learn sub-regions of the admissible state-space where replacing the strong solver with the weak solver maintains constraint satisfaction properties and does not result in a significant deterioration of performance. We illustrate the practical time-complexity reduction of the proposed solver selection mechanism on a station-keeping problem for a satellite.

 

  • Related Videos