Driving risk potential field modeling in heterogeneous traffic flow with applications to motorcycle car-following analysis

In the highway scenario, this study develops a heterogeneous traffic flow driving risk potential field model that accounts for both motorcycles and passenger cars. Considering the smaller size and higher maneuverability of motorcycles, we propose a motorcycle-specific driving risk potential field based on their geometric features and dynamic attributes. The model consists of a two-layer structure: (1) a scalar potential field that characterizes the spatial distribution of quantified risk; (2) a force field (vector field) derived from the potential gradients that describes the risk interaction and mutual influence among traffic participants. On this basis, we design a motorcycle trajectory prediction and simulation scheme grounded in the proposed model to analyze and investigate their car-following behavior in heterogeneous traffic flow. The results indicate that motorcycle car-following behavior is characterized by shorter following durations, smaller following distances, and narrower headway gaps. Due to their heightened risk perception and agile maneuverability, motorcycles exhibit relatively high motion noise during car-following maneuvers. The simulation results based on the proposed driving risk potential field effectively capture the interactions between vehicles, thereby achieving simulation of motorcycle motion states and driving trajectories with an average error of less than 1.8 m, demonstrating high accuracy and efficiency. These findings further confirm the significant potential of risk potential field modeling in real-time traffic risk perception and trajectory prediction, showing promise as a theoretical framework for autonomous driving in complex and mixed traffic environments.