Collaborative Research: WoU-MMA: Optimal Follow-up for Multimessenger Astronomy

Drexel University, Philadelphia PA

Investigators

Abstract

Mergers of black holes and neutron stars emit gravitational waves that can be detected on Earth by advanced detectors that sense minuscule disturbances in the fabric of the universe. In some cases, such events are accompanied by a burst of electromagnetic radiation that can also be seen by telescopes. These light signals combined with gravitational waves provide us with unprecedented insights into some of the most extreme objects in the Universe. However, the detection of such emission is challenging as it usually brief and faint. This project aims to develop cutting-edge artificial intelligence (AI) systems that will optimize the search for such sources among the hundreds of cosmic explosions that light up the night sky. Such systems will represent some of the first that make real-time scientific decisions in astronomy, determining the most efficient use of limited telescope resources to streamline the discovery process. This allows astronomers to focus on the scientific interpretation of results. The project will also train students in the advanced techniques required to design similar systems in other domains, such as robotics and finance. The investigators will develop a system that automates the follow-up decision-making step of the kilonova discovery infrastructure. Specifically, given an alert of gravitational wave mergers and gamma-ray bursts in the form of survey light curves and image stamps, and any value-added information, like galaxy redshift, the system will direct a series of resource assignments within a finite horizon that maximize a designated objective. The novel approach involves an AI agent that adaptively learns to make the best sequence of decisions given incomplete information and stochasticity concerning future survey and supplemental follow-up data from other sources. It will use the framework of reinforcement learning to rehearse gravitational wave trigger scenarios and learn how taking a certain action influences benefits achieved downstream, to solve for the optimal set of decisions that maximizes benefits given an unseen scenario. The AI agent will be trained to handle both photometric and spectroscopic follow-up and simultaneously maximize both kilonova discovery and inference objectives. This project addresses/advances the goals of the Windows on the Universe Big Idea. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.