This $749,963 Project Grant was awarded by the National Science Foundation (NSF) Division of Computing and Communication Foundations on October 1, 2023. The grant is funded under the NSF's Computer and Information Science and Engineering (CFDA #47.070) program, which supports investigator-initiated research and education across all areas of computing, communications, and information science and engineering.
The grant was awarded to Northeastern University, a private, non-profit research university in Boston, to develop a novel methodology for building safe multi-agent reinforcement learning (MARL) systems. The project combines expertise from formal methods (FM) and reinforcement learning (RL) to create "safety shields" that prevent safety violations and "safety coaches" that train agents to learn from mistakes, enabling RL systems to be used in safety-critical settings. Key deliverables include new formal methods, MARL techniques, and applications of model learning and abstraction refinement to the MARL domain. The project aims to transform the way RL systems are developed and deployed, with broader impacts including broadening participation in research and involving undergraduate students.