Speaker
Description
We present a fast time-domain simulator designed to reproduce the dynamical behavior of Fabry–Pérot cavities. The simulator implements a recursive field propagation formalism that explicitly accounts for multiple round trips within the cavity, enabling accurate modeling of transient phenomena such as ring-down, during resonance crossings. It supports step by step dynamic mirror motion and assures high computational efficiency.
A key feature of the framework is its integration with reinforcement learning environments, enabling the development and benchmarking of advanced control strategies for cavity lock acquisition. In particular, the tool has been already used to train deep reinforcement learning agents to stabilize cavities under nonlinear conditions where traditional linear control methods are less effective. Moreover, different configurations, such as multiple cavities in cascade, could be implemented thanks to the modular architecture, making it a valuable tool for simulating physically interesting scenarios. This framework finds a natural application in next-generation gravitational-wave detectors, where lock acquisition remains particularly challenging due to strong nonlinear dynamics and transient effects in high-finesse optical cavities.