Project Grant 2426340

This National Science Foundation (NSF) Computer and Information Science and Engineering (CISE) Federal Grant Program (CFDA 47.070) Project Grant award to Arizona State University (ASU) in the amount of $209,267 provides funding to develop new technologies to close the simulation-to-reality (sim-to-real) gap in reinforcement learning (RL). The 3-year project will pursue three key mechanisms: randomization, alignment, and derivation. The randomization approach will generate a diverse set of simulators to better represent the real-world environment, the alignment mechanism will make the simulator more closely match the real world, and the derivation approach will directly derive an optimal policy from real-world offline data without requiring a simulator. The research aims to advance RL techniques and improve the real-world applicability and generalization of RL systems. ASU may provide subawards to collaborating institutions as part of executing the project.