Project Grant 2419564

This National Science Foundation (NSF) Division of Electrical, Communications and Cyber Systems award, CFDA 47.041 Engineering, will provide $193,000 from September 1, 2024 to August 31, 2027 to New York University (NYU) to develop new theories and methodologies for safe reinforcement learning in domains such as robotics, autonomous driving, and power systems.

The key products and services to be delivered under this Project Grant include: 1) Formulating safety measures as general objectives beyond the standard cumulative form and developing solution approaches for this general formulation; 2) Considering both intrinsic uncertainty and model uncertainty to ensure the resulting policy performs well and satisfies a specified risk level in the real environment; 3) Bridging the gap between Bayesian reinforcement learning and safe reinforcement learning for continually improving models and policies while maintaining safety; 4) Developing near-optimal policy learning algorithms that adapt to piecewise non-stationary environments; and 5) Applying a rigorous simulation approach for policy evaluation to identify unexpected unsafe behaviors before they occur. This research aims to make significant contributions to the field of safe reinforcement learning with broad applicability in various domains utilizing this technology.