Reinforcement learning assisted non-reciprocal optomechanical gyroscope (2503.08319v1)

Abstract: We propose a novel optomechanical gyroscope architecture based on a spinning cavity optomechanical resonator (COM) evanescently coupled to a tapered optical fiber without relying on costly quantum light sources. Our study reveals a striking dependence of the gyroscope's sensitivity on the propagation direction of the driving optical field, manifesting robust quantum non-reciprocal behavior. This non-reciprocity significantly enhances the precision of angular velocity estimation, offering a unique advantage over conventional gyroscopic systems. Furthermore, we demonstrate that the operational range of this non-reciprocal gyroscope is fundamentally governed by the frequency of the pumping optical field, enabling localized sensitivity to angular velocity. Leveraging the adaptive capabilities of reinforcement learning (RL), we optimize the gyroscope's sensitivity within a targeted angular velocity range, achieving unprecedented levels of precision. These results highlight the transformative potential of RL in advancing high-resolution, miniaturized optomechanical gyroscopes, opening new avenues for next-generation inertial sensing technologies.