CAREER: Multi-scale Multi-population Mean Field Game-Theoretic Framework for the Autonomous Mobility Ecosystem award

This Faculty Early Career Development (CAREER) grant will contribute to the improved well-being of individuals and increased U.S. economic competitiveness by assisting in the development of control methods for autonomous vehicles (AV). AVs are anticipated to improve traffic safety and efficiency. In the near future, however, AVs will operate on public roads in mixed traffic and will have to manage complex interactions with human-driven vehicles (HV). This award supports research that will lead to control paradigms for AVs operating in mixed traffic conditions, particularly when traffic is dense and safe operations require effective automated car-following and lane-changing controls. The project is expected to contribute to a better understanding of the future transportation ecosystem and the controls needed to guide the ecosystem toward an equilibrium that benefits society. The accompanying educational plan aims to fundamentally redesign the transportation engineering curricula via new graduate course development and outreach programs, leveraging the COSMOS testbed deployed in Columbia’s neighborhood. The outcomes of this research will be assessed by an advisory committee of select leaders from academia, public agencies, and the AV industry.

This research develops a new modeling framework that builds from the fields of game theory, dynamic control, data science, and transportation engineering. Mean-field game-theoretic methods are used to characterize the dynamic behavior of the mixed traffic system and to examine optimal policies associated with infrastructure planning and the regulation of technology. This framework provides a rigorous foundation for the development of a multi-agent simulation platform to inform policy and practice as part of the development of the transportation ecosystem. The analytical framework leverages the state-of-the-art techniques from game theory and AI methods. The research addresses an important gap in the autonomous driving control literature in which AVs are essentially modelled as human drivers that can 'react' faster, 'see' farther, and 'know' the road environment better.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.