AI-Assisted Sensing & Control at Gravitational Wave Detectors

Albert Einstein’s theory of relativity predicted gravitational waves as ripples in space-time. These waves can help us measure properties such as distance, mass, and the spin of astronomical objects, including neutron stars and black holes. Despite Einstein's skepticism, scientists have made significant technological advancements over the last century that have now made it possible to detect gravitational waves.

In his talk, Nikhil Mukund from LIGO Labs at the Massachusetts Institute of Technology (MIT) dives into the groundbreaking advancements in gravitational wave detection and the application of artificial intelligence, which enabled scientists to identify disturbances as small as one ten-thousandth the diameter of a proton.

Laser interferometers divide laser beams into two parts, which travel several kilometers before bouncing off two mirrors and then recombine at the beam splitter, where information is extracted from any disturbances in the laser beams' paths. These devices are extremely sensitive and are prone to disturbances from the environment, such as ground vibration from urban traffic, machinery, sea tides, and seismic activity. The talk explores sophisticated control systems and techniques to maintain optimal performance, a challenging task given the myriads of subsystems involved. These complexities give scientists an opportunity to apply artificial intelligence techniques to filter noise from the true signals. By using reinforcement learning and introducing AI agents, the team was able to address nonlinear dynamics and cross-couplings inherent in the system, tackling tasks such as vibration isolation, laser beam alignment, and thermal correction.