Lokesh Muruga | Searcharxiv

arXiv2026

Work Extraction via Backward Motion in Optimal Closed-Loop Stochastic Control

We experimentally realize finite-time feedback control in an overdamped colloidal system using real-time optical tweezers with in situ reinforcement learning (RL). By varying the protocol duration tf for displacing the optical trap between prescribed positions, the optimal strategies identified by RL reveal a crossover from deterministic dragging toward the target to feedback-assisted exploitation of thermal fluctuations, reducing and eventually overcoming the energetic cost. The resulting policies agree quantitatively with the exact optimal closed-loop solution. By extending the approach to spatially localized external forcing, we further show that RL can identify optimal feedback strategies in heterogeneous stochastic environments where direct analytical control design is challenging.

cond-mat.stat-mech↗

arXiv2026

Extracting work from hidden degrees of freedom

Thermodynamics establishes that information acquired through measurement can be converted into work, as exemplified by Maxwell's demon and Szilard engines. Most experimental realizations of information engines, however, implicitly assume Markovian environments, in which information exchanged with the surroundings is irreversibly lost. Many physical systems instead exhibit environmental memory, with hidden degrees of freedom retaining correlations with the system's past and giving rise to non Markovian dynamics. Whether and how such concealed memory can be harnessed as a thermodynamic resource has remained an open question. Here we experimentally demonstrate work extraction from environmental memory. Using time resolved measurements on an optically trapped Brownian particle in equilibrium, we implement a time delayed double measurement protocol that retrieves information via backflow from hidden bath degrees of freedom. We show that this information backflow alters relaxation dynamics, can be quantified independently of initial state effects, and when appropriately exploited enhances work extraction. Notably, we identify regimes in which the extracted work exceeds the energy stored in the observable degree of freedom alone. Our results establish environmental memory as an experimentally accessible thermodynamic resource and reveal how non Markovian dynamics can be systematically explored to improve the performance of information engines operating in time-correlated environments.

cond-mat.soft↗