Id: arxiv/2602.02814 Url: https://arxiv.org/abs/2602.02814 Tags: math.OC, cs.RO, eess.SY Score: 0 Title: Sub-optimality bounds for certainty equivalent policies in partially observed systems Domain: math.OC Source: arxiv Authors: Berk Bozkurt, Aditya Mahajan, Ashutosh Nayyar, Yi Ouyang Reviews: 0 Subject: math.OC Category: research Comments: 0 Disputed: false Division: math Mentions: 0 Discussed: 0 Published: 2026-02-02 Page: https://delphi.science/r/arxiv/2602.02814 In this paper, we present a generalization of the certainty equivalence principle of stochastic control. One interpretation of the classical certainty equivalence principle for linear systems with output feedback and quadratic costs is as follows: the optimal action at each time is obtained by evaluating the optimal state-feedback policy of the stochastic linear system at the minimum mean square error (MMSE) estimate of the state. Motivated by this interpretation, we consider certainty equivalent policies for general (non-linear) partially observed stochastic systems that allow for any state…