Randomized Optimal Control Problem via Reinforcement Learning Under Knightian Uncertainty: Presenting a Unified Analytical Framework
IEEE Systems, Man, and Cybernetics Magazine · 2026-07 · 12 (3) · 102-115
DOI: 10.1109/msmc.2026.3659306 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Systems, Man, and Cybernetics Magazine, 2026-07. Verfasst von Ziyu Li und 2 weiteren Autor:innen. Die Arbeit nennt 36 Literaturangaben. Gefördert von: National Natural Science Foundation of China.
Abstract (Originalsprache)
Considering that the decision-making environment faced by reinforcement learning (RL) agents is full of Knightian uncertainty, we describe the exploratory state dynamics equation to study the entropy-regularized relaxed stochastic control problem in a Knightian uncertainty environment. By employing stochastic analysis theory and the dynamic programming principle under nonlinear expectation, we derive the Hamilton-Jacobi-Bellman (HJB) equation and solve for the optimal policy that achieves a tradeoff between exploration and exploitation. Subsequently, for the linear-quadratic (LQ) case, we prove that the optimal randomized feedback control follows a Gaussian distribution. Furthermore, we investigate how the degree of Knightian uncertainty affects the variance of the optimal feedback policy. Finally, we establish the solvability equivalence between nonexploratory and exploratory LQ problems under Knightian uncertainty and analyze the associated exploration cost.
Förderung
- National Natural Science Foundation of China
Zitieren
APA
Li, Z., Fei, C., & Fei, W. (2026). Randomized Optimal Control Problem via Reinforcement Learning Under Knightian Uncertainty: Presenting a Unified Analytical Framework. IEEE Systems, Man, and Cybernetics Magazine, 12(3), 102-115. https://doi.org/10.1109/msmc.2026.3659306
Vancouver
Li Z, Fei C, Fei W. Randomized Optimal Control Problem via Reinforcement Learning Under Knightian Uncertainty: Presenting a Unified Analytical Framework. IEEE Systems, Man, and Cybernetics Magazine. 2026;12(3):102-115. doi:10.1109/msmc.2026.3659306
BibTeX
@article{Li2026,
author = {Li, Ziyu and Fei, Chen and Fei, Weiyin},
title = {Randomized Optimal Control Problem via Reinforcement Learning Under Knightian Uncertainty: Presenting a Unified Analytical Framework},
journal = {IEEE Systems, Man, and Cybernetics Magazine},
year = {2026},
volume = {12},
number = {3},
pages = {102--115},
doi = {10.1109/msmc.2026.3659306}
}
EUVIMED