Constructing optimal feedback loops in stochastic thermodynamic systems

Develop general methods to construct optimal feedback control policies for stochastic thermodynamic systems with feedback, including mesoscopic information engines based on colloidal particles in optical traps, even when measurement costs are not modeled explicitly.

Background

The paper studies optimal control of an overdamped Brownian particle in a harmonic optical trap, formulating the problem as a discrete-time POMDP and leveraging LQG control to decouple measurement scheduling from physical trap control. Prior literature has focused mainly on optimizing trap trajectories under fixed measurement architectures, and exact analytical solutions have been scarce due to the complexity of continuous-time partial observability and associated Hamilton-Jacobi-Bellman equations.

Within this broader context, the authors explicitly note that, in general, constructing optimal feedback loops in systems governed by stochastic thermodynamics remains an open problem, even in simplified settings where measurement costs are not considered. Their work addresses a specific instance with costly measurements but does not resolve the general design question across all feedback-controlled stochastic systems.