This $350,000 Project Grant awarded by the National Science Foundation's Computer and Information Science and Engineering (CISE) program (CFDA 47.070) will support the development of an efficient human-in-the-loop learning framework for human-centric cyber-physical systems (CPS). The key products and services to be delivered include:
-
A novel human-in-the-loop learning approach that incorporates a human subject to guard the learning agent, enabling the human to actively intervene in unsafe situations and demonstrate correct actions to the agent during training. This reward-free learning method will encourage efficiency, safety, and AI alignment.
-
The creation of digital twins of task-specific human behaviors to evaluate the proposed learning method in targeted CPS applications, with a focus on simulated environments for human behaviors in driving and exoskeleton use cases.
-
The empirical evaluation and real-world experimentation of the proposed learning methods in the targeted CPS applications across various simulated and actual scenarios to validate their performance.
The project aims to facilitate harmonious human-machine collaborations and enable efficient and safe autonomy in CPS, with potential impacts across industries such as automotive, package delivery, and rehabilitation. No sub-awards are planned for this award, which has an ultimate completion date of September 30, 2027.
Generated 2/25/25, 3:08 AM