Mirror Descent for Deterministic Optimal Control
Abstract
We study an explicit mirror-descent method for finite-horizon deterministic optimal control problems. The method is motivated by Pontryagin's maximum principle: at each iteration, one solves the state and adjoint equations and updates the control by maximizing a first-order approximation of the regularized Hamiltonian penalized by a Bregman divergence. In the Euclidean case, the update reduces to a projected gradient step in the control variable. Under global smoothness assumptions and uniform convexity of the mirror map, we prove a relative smoothness estimate for the cost functional and derive an energy dissipation inequality for sufficiently small step sizes. Under an additional concavity assumption on the unregularized Hamiltonian and convexity of the terminal cost, we establish relative convexity of the regularized objective. These estimates yield an O(1/n) convergence rate in the unregularized convex case and a geometric rate when the control regularization parameter is positive. Numerical examples illustrate the behavior of the method in linear-quadratic, degenerate convex, and nonlinear high-dimensional settings.
Related
- A stochastic mirror descent algorithm for mean-field control problems
- Does Policy Mirror Descent Converge in Policy with Non-Decomposable Mirror Maps? A Counterexample
- Mirror descent algorithms with logarithmic barriers
- Stochastic Mirror Descent under Dependent Observations
- An analysis of Mirror-Descent Soft Actor-Critic
BibTeX
@misc{feng2026mirror,
title = {{Mirror Descent for Deterministic Optimal Control}},
author = {Ye Feng and Jianfeng Lu},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2605.02653},
eprint = {2605.02653},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2605.02653},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Mirror Descent for Deterministic Optimal Control
AU - Ye Feng
AU - Jianfeng Lu
PY - 2026
DA - 2026/09/02
PB - arXiv
DO - 10.48550/arxiv.2605.02653
UR - https://arxiv.org/abs/2605.02653
AB - We study an explicit mirror-descent method for finite-horizon deterministic optimal control problems. The method is motivated by Pontryagin's maximum principle: at each iteration, one solves the state and adjoint equations and updates the control by maximizing a first-order approximation of the regularized Hamiltonian penalized by a Bregman divergence. In the Euclidean case, the update reduces to a projected gradient step in the control variable. Under global smoothness assumptions and uniform convexity of the mirror map, we prove a relative smoothness estimate for the cost functional and derive an energy dissipation inequality for sufficiently small step sizes. Under an additional concavity assumption on the unregularized Hamiltonian and convexity of the terminal cost, we establish relative convexity of the regularized objective. These estimates yield an O(1/n) convergence rate in the unregularized convex case and a geometric rate when the control regularization parameter is positive. Numerical examples illustrate the behavior of the method in linear-quadratic, degenerate convex, and nonlinear high-dimensional settings.
ER -