First AI-driven telescope goes stargazing

• For Journalists
• News Releases
• Latest Releases
• News Release

First AI-driven telescope goes stargazing

New AI system successfully schedules observations on a major national telescope

• Planning telescope observations requires balancing ever-changing conditions
• Telescope time is scarce, making every observing decision count
• AI learned to schedule observations from 13 years of historical telescope data
• It generated an observing plan and adapted in real time as conditions changed

EVANSTON, Ill. - Every night, astronomers must carefully assess changing weather, the intensity of moonlight and shifting atmospheric conditions before deciding where to point a telescope. It's a constant balancing act designed to squeeze as much science as possible from every precious hour beneath dark skies.

Now, Northwestern University, University of Chicago and Fermilab scientists have developed a new artificial intelligence (AI) tool that automatically determines where a telescope should point.

After developing the tool in the National Science Foundation (NSF)-Simons Foundation AI Institute for the Sky (SkAI, pronounced "sky"), the scientists successfully used the AI system to schedule observations with the 570-megapixel U.S. Department of Energy (DOE)-fabricated Dark Energy Camera (DECam), mounted on the NSF Víctor M. Blanco 4-meter Telescope at Cerro Tololo Inter-American Observatory (CTIO) in Chile. Not only did the system generate an observing plan, but it also adapted that plan in real time as environmental conditions changed. By automating routine scheduling decisions, the innovation will help telescopes collect the best possible data.

"This is an important milestone toward more autonomous observatories," said UChicago's Alex Drlica-Wagner, who co-led the project. "One of the main achievements is that we set up all the infrastructure needed to deploy this self-driving telescope on a national observatory. Currently, I would say its performance is comparable to a human's ability. As the next step, we plan to teach the computer to do a better job than a human."

Drlica-Wagner is a scientist at Fermilab and a professor of astronomy and astrophysics at UChicago. He co-led the project with Aravindan Vijayaraghavan, an associate professor of computer science at Northwestern's McCormick School of Engineering. Paul Chichura, SkAI postdoctoral associate; Rachel Hur, a Ph.D. student at UChicago; and Guillermo Damke, an associate scientist at NSF's NOIRLab; performed the on-sky deployment at the Blanco telescope. Drlica-Wagner, Vijayaraghavan, Chichura and Hur are core members of SkAI.

Choosing where to point a telescope isn't just about finding an interesting object to observe. It's also about making the most of every minute of valuable observing time. Sometimes, astronomers wait months for a chance to use a major telescope. A poorly positioned telescope could return less-sharp images or washed-out images flooded by moonlight, making faint or distant objects even more difficult to detect. And the opportunity to redo the failed observation might be months away.

"Large telescopes are national or international resources," Drlica-Wagner said. "Many astronomers around the world want time to use these telescopes, and that time is limited. If everyone could use their time more efficiently, then the community will be able to do more science."

To make the observing process more efficient, Drlica-Wagner combined his expertise in large astronomical surveys with Vijayaraghavan's expertise in cutting-edge machine learning. With their teams at SkAI, they developed a deep-learning scheduling system. Rather than programming AI with rules astronomers have developed over decades, the researchers let the system learn on its own. They trained a deep-learning model on historical observations from the DOE-funded Dark Energy Survey, which scans the night sky with a giant camera mounted on the Blanco Telescope.

"We trained the model on years of historical observations by showing it where the telescope was pointing at one moment and asking it to predict the next observation," Drlica-Wagner said. "Then we compared its prediction to what astronomers actually did and asked it to correct its mistakes. After repeating this process many times, it learned how to schedule observations without being explicitly taught how the brightness of the moon, the atmospheric conditions or the many other factors affect the quality of astronomical observations."

"It is exciting to see ideas from AI and reinforcement learning brought to telescope scheduling, where every decision must balance changing conditions and scarce observing time," Vijayaraghavan said. "Developing intelligent scheduling systems for astronomical surveys also raises fascinating new machine learning problems, and we are excited to continue exploring them through this project."

This past spring and summer, the intelligent scheduling system completed two successful observing runs on the Blanco Telescope - one of the world's most productive astronomical facilities. For this initial deployment, the goal was to get the AI to perform about as well as human schedulers. The team's next goal is to teach the AI not just to mimic human decision-making but improve upon it. By exploring observing strategies humans might never consider, AI eventually could make telescopes even more efficient.

As next-generation telescopes including the NSF-DOE Vera C. Rubin Observatory begin producing unprecedented amounts of astronomical data, intelligent scheduling systems could help companion telescopes respond more efficiently and maximize the scientific value of every observation run.

"If we can automate this technical operational task so it requires less human effort, then astronomers can have more time to think about more scientifically interesting problems and focus on discovery," Drlica-Wagner said.

Led by Northwestern University, SkAI is a National AI Research Institute jointly funded by the NSF and the Simons Foundation. SkAI brings together researchers in astronomy, AI and related fields to develop trustworthy AI tools that accelerate scientific discovery, advance cutting-edge astronomical surveys and instruments, and train the next generation of interdisciplinary scientists.

Embed Video:

NSF Víctor M. Blanco 4-meter Telescope

Aravindan Vijayaraghavan

Co-lead

Associate professor of computer science

• [email protected]