{"id":"arxiv/2608.11808","url":"https://arxiv.org/abs/2608.11808","tags":["math.OC"],"score":0,"title":"Data-Driven optimal control via Koopman operators and Hamilton-Jacobi-Bellman equations","domain":"math.OC","source":"arxiv","authors":["Guuyuan Chen","Gaosheng Zhu"],"reviews":0,"subject":"math.OC","summary":"This paper presents a data-driven stable manifold (DD-SM) method, which integrates Koopman operator representation learning with the geometric stable manifold approach to Hamilton-Jacobi-Bellman (HJB) equations, enabling end-to-end optimal feedback control synthesis from raw trajectory data without prior knowledge of system dynamics. We construct an augmented control system under a unified symmetric subspace decomposition (SSD) and extended dynamic mode decomposition (EDMD) framework for joint approximation of the drift field, control matrix and their spatial derivatives, and derive…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-08-12","reactions":{}}