Separable approximations of optimal value functions under a decaying sensitivity assumption
arXiv:2304.06379 · doi:10.1109/CDC49753.2023.10383497
Abstract
An efficient approach for the construction of separable approximations of optimal value functions from interconnected optimal control problems is presented. The approach is based on assuming decaying sensitivities between subsystems, enabling a curse-of-dimensionality free approximation, for instance by deep neural networks.
References in corpus (5)
- Adaptive Deep Learning for High-Dimensional Hamilton-Jacobi-Bellman Equations
- Exponential Decay of Sensitivity in Graph-Structured Nonlinear Programs
- Near-Optimal Distributed Linear-Quadratic Regulator for Networked Systems
- Controllability and Observability Imply Exponential Decay of Sensitivity in Dynamic Optimization
- Separable approximations of optimal value functions under a decaying sensitivity assumption