{"id":"arxiv/2608.07433","url":"https://arxiv.org/abs/2608.07433","tags":["math.OC","cs.LG"],"score":0,"title":"Wasserstein Policy Gradient for Entropy-Regularized Linear-Quadratic Control","domain":"math.OC","source":"arxiv","authors":["Zhaoyu Zhu","Rui Gao","Shuang Li"],"reviews":0,"subject":"math.OC","summary":"Wasserstein policy gradient (WPG) updates state-conditional action laws by transport in the action space. We study entropy-regularized discounted linear-quadratic (LQ) control. A Bellman verification argument shows that the unrestricted problem has a linear-Gaussian optimal policy, and the discounted-occupancy-weighted statewise Wasserstein gradient is tangent to this policy class. WPG therefore reduces exactly to a finite-dimensional ODE for the feedback gain and action covariance. We prove that this ODE is globally well posed and converges exponentially from every admissible…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-08-07","reactions":{}}