The Laser Interferometer Gravitational-Wave Observatory (LIGO) has achieved phenomenal successes in recent years with its new gravitational-wave detection. That progress is made possible in large part by a new artificial intelligence algorithm developed in collaboration with Google DeepMind. While we’re thrilled by this development, it mostly advances the cause of reducing annoying noise. In particular, it aims for LIGO’s lower-frequency band, which is critical in the detection of more massive black hole mergers.
LIGO is the scientific collaborative behind two such facilities, each shaped like gigantic “L’s,” spread out across Louisiana and Washington. Each arm of LIGO’s gigantic structure contains a vacuum tube in which lie super-sensitive laser technology. These lasers travel back and forth through 4-kilometer-long tubes, stabilized with 40-kilogram mirrors hanging by wires at each end. Gravitational waves from remote corners of space travel directly to Earth. As they come in, one arm of LIGO gets longer or shorter by minuscule distances, allowing scientists to see these hard-to-find cosmic happenings.
LIGO has struggled with an issue called controller-induced hiss. This nuisance noise presents a serious challenge, contaminating valuable gravitational-wave signals, including the noise-critical 10 to 30 Hertz range. This range is vital for capturing data on massive black hole mergers and observing black holes nearing their final death spirals.
The Role of AI in Noise Reduction
To combat the interference of this hiss, the scientists at LIGO decided to employ the help of artificial intelligence. This exciting new technique—which they call Deep Loop Shaping—uses the powerful AI technology to more effectively control the vibrations in LIGO’s mirrors. This new innovation doubles LIGO’s performance overnight. It also heralds unprecedented promise for applications across mechanical, civil, aerospace and other engineering disciplines requiring vibration suppression and noise cancellation.
Rana Adhikari, a physicist at Caltech, is one of the project’s leaders. He underscored the profound societal implications technology like this can have. “This technology will help us not only improve LIGO but to build next-generation, even bigger gravitational-wave detectors,” he stated.
Deep Loop Shaping is subjected to a strict adversarial training regimen. With each round of rich data, Google DeepMind conducts thousands of simulations to fine-tune the algorithm’s predictive power. Adhikari explained further on this, noting that, “we provided the training data, and Google DeepMind conducted the simulations. In short, they were operating dozens of simulated LIGOs at once. You might imagine the training as a game. You score more points the lower the amount of noise is and are penalized if you increase the noise. The best ‘players’ win, and they continue playing in order to win the game of LIGO. The result is stunning—the algorithm is able to effectively suppress mirror noise.
Implications for Future Research
The effects of Deep Loop Shaping go even further than advancing LIGO’s abilities. It paves Innovative Agritech corridors, enabling cutting-edge research in aerospace, robotics, and structural engineering among others. Brendan Tracey and Jonas Buchli, who are part of the research team, noted, “In the future, Deep Loop Shaping could be applied to many other engineering problems involving vibration suppression, noise cancellation and highly dynamic or unstable systems.”
An expert in control systems, Richard Murray saw the promise of AI. He emphasized the advantages it holds for unveiling patterns in intricate realms that previous analytical approaches have overlooked. “With AI, if you train it on a model of sufficient detail, it can exploit features in the system that you wouldn’t have considered using classical methods,” he explained.
Even if gravitational-wave technologies get applied in different areas, their application might encourage the next generation to participate in gravitational-wave work. Adhikari expressed hope that advancements like these would motivate students to join the field: “We think this research will inspire more students to want to work at LIGO and be part of this remarkable innovation.”
Enhancing Detection Capabilities
Above all, LIGO is a place that is always improving and developing its technology. Our researchers are equally thrilled at the prospects to detect larger black holes and more cosmic events than ever before! The partnership with Google DeepMind represents a huge step forward for LIGO. Jointly, they will improve the hunt for gravitational waves and further expand our understanding of the cosmos.
Christopher Wipf, another researcher as part of the project, analogized the noise-canceling mechanism utilized in LIGO to that of headphones. “Imagine you are sitting on the beach with noise-canceling headphones. A microphone picks up the ocean sounds, and then a controller sends a signal to your speaker to counteract the wave noise,” he said. “This is similar to how we control ocean and other seismic ground-shaking noise at LIGO.”
Wipf further elaborated on the challenges posed by self-inflicted noise: “If you have ever listened to these headphones in a quiet area, you might hear a faint hiss. The microphone has its own intrinsic noise. This self-inflicted noise is what we want to get rid of in LIGO.”
With the introduction of Deep Loop Shaping, this enormously complex problem becomes far less intimidating. Adhikari stated, “We were already at the forefront of innovation, making the most precise measurements in the world, but with AI, we can boost LIGO’s performance to detect bigger black holes.”