Neural Solvers for Fast and Accurate Numerical Optimal ControlDownload PDF

29 Sept 2021, 00:33 (edited 13 Mar 2022)ICLR 2022 PosterReaders: Everyone
  • Keywords: Deep Learning, Numerical Methods, Optimal Control
  • Abstract: Synthesizing optimal controllers for dynamical systems often involves solving optimization problems with hard real-time constraints. These constraints determine the class of numerical methods that can be applied: computationally expensive but accurate numerical routines are replaced by fast and inaccurate methods, trading inference time for solution accuracy. This paper provides techniques to improve the quality of optimized control policies given a fixed computational budget. We achieve the above via a hypersolvers approach, which hybridizes a differential equation solver and a neural network. The performance is evaluated in direct and receding-horizon optimal control tasks in both low and high dimensions, where the proposed approach shows consistent Pareto improvements in solution accuracy and control performance.
  • One-sentence Summary: We propose a hypersolvers approach for numerical optimal control which shows consistent Pareto improvements in solution accuracy and control performance.
  • Supplementary Material: zip
12 Replies

Loading