A collaboration led by researchers at the U.S. Department of Energy's Argonne National Laboratory, working with Johns Hopkins University, the University of Chicago and the University of Illinois Urbana-Champaign, has developed a new AI-powered framework to tackle this problem. The system, called RADAR, for Radio Afterglow Detection and AI-driven Response, is designed to combine gravitational wave data with radio observations in a coordinated, resource-aware way.

Multi-messenger astrophysics moved into a new phase in 2017 with GW170817, the first neutron star merger detected in both gravitational waves and light. By uniting information from gravitational waves and electromagnetic radiation, scientists were able to measure the properties of the merger, probe its surroundings and study how some heavy elements are synthesized, revealing aspects of the event that no single messenger could provide on its own.

Realizing such comprehensive follow-up in practice is difficult. Gravitational wave alerts can initially span large regions of the sky, observatories must react quickly, and telescope time and computing capacity are limited and highly competitive. Coordinating observations and analysis in real time, particularly at radio wavelengths where emission is often faint and delayed, has become one of the main bottlenecks for multi-messenger studies.

Radio telescopes typically observe small patches of sky, and the radio afterglows from neutron star mergers and similar events can take months or even years to emerge at detectable levels. As new generations of detectors come online, astronomers expect to identify hundreds to thousands of gravitational wave sources annually, far beyond what traditional, manual follow-up strategies can handle.

RADAR addresses these pressures by bringing artificial intelligence and high-performance computing to bear directly where the data reside. The framework runs at supercomputing centers, analyzes gravitational wave and radio datasets in situ and minimizes the need to transfer large volumes of information between sites. It is built to respect data-access controls, including proprietary restrictions, while still enabling rapid alerts and joint analysis.

When RADAR identifies a gravitational wave event, it automatically initiates searches for associated radio signals. The system uses large language models to read and interpret public notices and telegrams from observatories worldwide, extracting relevant details to inform follow-up strategies. Crucially, RADAR can operate on private radio data without exposing raw datasets, allowing teams to collaborate while preserving data rights.

"This framework shows how we can do collaborative, cutting-edge astrophysics while respecting data rights and privacy," said Eliu Huerta, a theoretical physicist at Argonne, the University of Chicago and the University of Illinois Urbana-Champaign. "RADAR is built to grow with the field, ensuring we can meet the challenges of the multi-messenger era."

The team evaluated RADAR using GW170817, the only neutron star merger so far with a confirmed electromagnetic counterpart. In those tests, RADAR successfully combined gravitational wave measurements with both public and private radio observations, refining estimates of the event's geometry and distance. The joint analysis demonstrated how coordinated use of gravitational wave and radio data can guide more effective follow-up campaigns and accelerate discovery.

"Multi-messenger astronomy thrives on coordination," said Alessandra Corsi from Johns Hopkins. "RADAR gives us a way to plan and adapt follow-up strategies, even when the data itself can't be shared directly. This capability will become increasingly critical as next-generation detectors transform today's trickle of multi-messenger detections into a flood."

Developing RADAR required close cooperation among astrophysicists, AI specialists and computing engineers. The framework was exercised end-to-end at several leading computing facilities using real datasets from GW170817, demonstrating that it can reproduce the main results of the original multi-messenger analysis while distributing work across heterogeneous systems.

Campaigns on resources including the Polaris supercomputer at the Argonne Leadership Computing Facility, Delta and DeltaAI at the National Center for Supercomputing Applications, and Advanced Research Computing at Johns Hopkins showed that RADAR can reduce data movement, uphold access restrictions and still orchestrate large-scale analyses reliably. The Argonne Leadership Computing Facility is a DOE Office of Science user facility dedicated to open science.

"Developing the AI models for gravitational wave detection was exciting because we could see them working in real time across different supercomputers," said Victoria Tiki of the University of Illinois Urbana-Champaign and Argonne. "That speed and adaptability are crucial for the next generation of gravitational wave events."

"From an engineering perspective, building RADAR meant integrating AI, federated computing and high-performance infrastructure in a way that's seamless for scientists," added Parth Patel from Argonne. "It's about turning cutting-edge technology into practical tools for discovery."

One of RADAR's distinctive features is its use of AI to read the human-generated messages that astronomers rely on to share results from telescopes around the globe. By automatically parsing these notices, the system can keep track of evolving observations and adjust follow-up plans dynamically.

"One particularly exciting aspect of this project was integrating large language models into RADAR to automate the processing of notices and telegrams," said Kara Merfeld from Johns Hopkins. "This work demonstrated both the potential of this approach and the opportunities for further optimization."

Looking ahead, the RADAR team plans to extend the framework's capabilities. They are working on AI models that can forecast gravitational wave events before they are fully observed, giving astronomers more time to position telescopes and schedule observations in advance.

The researchers also aim to accelerate radio data modeling so that signals can be interpreted more quickly. As new gravitational wave and radio facilities begin operations, RADAR's flexible architecture is expected to support fast, efficient follow-up campaigns for a broader array of cosmic messengers, potentially including neutrinos and other elusive signals from space.

The results of the RADAR study appear in The Astrophysical Journal Supplement Series, providing a detailed account of the framework and its performance on GW170817. The work involved contributors from Argonne, the University of Chicago, Johns Hopkins University and the University of Illinois Urbana-Champaign, and received support from the U.S. Department of Energy and the National Science Foundation.

Research Report:Radio Afterglow Detection and AI-driven Response (RADAR): A Federated Framework for Gravitational-wave Event Follow-up

Related Links
Argonne National Laboratory
The Physics of Time and Space