Optimal Bias Potentials via Ergodic Optimal Control and Generator LearningNew

This paper presents a method for computing optimal bias potentials to accelerate transitions between metastable states in molecular dynamics simulations. The authors formulate optimal biasing as an ergodic optimal control problem, which can be equivalently expressed as a linear eigenvalue problem for the infinitesimal generator of the unbiased dynamics. Data-driven learning methods are employed to learn the generator, enabling the computation of bias potentials without requiring full knowledge of the system dynamics. The approach is applicable to equilibrium property computation in materials science and computational chemistry simulations.