English

Trajectory-Regularized Stochastic Optimal Control via KL Divergence

Systems and Control 2026-07-24 v1 Machine Learning

Abstract

We introduce trajectory-regularized stochastic optimal control (TRSOC), which augments standard stochastic optimal control (SOC) with a Kullback--Leibler (KL) divergence between controlled and reference trajectory distributions. Using Girsanov's theorem, the trajectory KL reduces to a quadratic drift mismatch penalty, yielding a modified running cost that preserves the dynamic programming (DP) structure. We derive the corresponding Hamilton--Jacobi--Bellman (HJB) equation and characterize the optimal policy. In the linear-quadratic (LQ) setting, the formulation admits a closed-form solution with an augmented control cost. Experiments show that the regularization parameter induces a trade-off between performance-driven and reference-preserving behavior, including cases with reference dynamics learned from offline data.

Cite

@article{arxiv.2607.22201,
  title  = {Trajectory-Regularized Stochastic Optimal Control via KL Divergence},
  author = {Mintae Kim and Koushil Sreenath},
  journal= {arXiv preprint arXiv:2607.22201},
  year   = {2026}
}

Comments

8 pages, 4 figures, 65th IEEE Conference on Decision and Control