English

Learning algorithms for mean field optimal control

Optimization and Control 2025-03-25 v1

Abstract

We analyze an algorithm to numerically solve the mean-field optimal control problems by approximating the optimal feedback controls using neural networks with problem specific architectures. We approximate the model by an NN-particle system and leverage the exchangeability of the particles to obtain substantial computational efficiency. In addition to several numerical examples, a convergence analysis is provided. We also developed a universal approximation theorem on Wasserstein spaces.

Keywords

Cite

@article{arxiv.2503.17869,
  title  = {Learning algorithms for mean field optimal control},
  author = {H. Mete Soner and Josef Teichmann and Qinxin Yan},
  journal= {arXiv preprint arXiv:2503.17869},
  year   = {2025}
}