{"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","summary":"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…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-02-02","reactions":{}}