LIGO Document T2600198-v2
- This document proposes the development of an accurate, dynamically updatable digital twin of the Caltech 40m Prototype Input Mode Cleaner (IMC) to serve as an offline testing and training environment. We utilize a gray-box system identification approach to model the suspended optics and triangular cavity, tuned with and validated against empirical hardware data. A digital twin is compiled using the rtsfreerun framework, ensuring compatibility with LIGO s real-time control architecture. Finally, a validated digital twin distribution serves as the simulated plant to train a reinforcement learning policy. Through deployment of the resulting policy, this project aims to reduce technical noise and thereby contribute to the sensitivity targets of future upgrades and next-generation detectors.
DCC Version 3.6.0, contact
DCC Help