Driving behavior modeling & prediction
With the integration of Autonomous Vehicles (AVs) and CAVs, the dynamics of mixed traffic flows have shifted significantly compared to traditional traffic conditions. In mixed traffic, human-driven vehicle (HV) behaviors exhibit greater variability and uncertainty, particularly in longitudinal car-following and lateral maneuvers. These factors heighten traffic safety risks and complicate the decision-making processes of AVs. To address these challenges, we have developed novel modeling and prediction methodologies for both AVs and HVs, focusing on longitudinal and lateral driving behaviors. These methodologies leverage approaches including physics-based models, machine learning (ML), physics-informed machine learning (PIML), and foundation models (FMs)/large language models (LLMs).