Modeling risk potential fields for mandatory lane changes in intelligent connected vehicle environment

Abstract

In the environment of Intelligent Connected Vehicles (ICVs), the ICV system is capable of gathering, processing, and transmitting the majority of traffic information to each vehicle, thereby empowering drivers to make more informed and secure decisions. Nevertheless, ICVs are still incapable of completely avoiding traffic conflicts and accidents. Consequently, developing a risk assessment model for ICVs is crucial for their deployment. To investigate the dynamic patterns of lane-changing vehicle risks within ICV environments, this study developed a quantitative relationship model based on the Artificial Potential Field (APF) theory. The model correlates the Total Potential Field (TPF) of mandatory lane-changing (MLC) vehicle with the longitudinal/transverse gravitational potential field (LGPF/TGPF) and the longitudinal/transverse repulsion potential field (LRPF/TRPF). Subsequently, three sets of Human Machine Interface (HMI) systems were developed to delineate distinct ICV environments: baseline, warning group, and guidance group. The baseline furnishes fundamental data, the warning group supplements preceding vehicle icons and real-time headway information, whereas the guidance group additionally augments speed and voice guidance functionalities. Following this, a simulated driving experiment involving 43 participants was carried out to simulate MLC scenarios in highway merging sections. Using enumeration and gradient descent methods, the quantitative relationship between different types of risk fields was determined. The findings indicated that the ICV environment significantly impacts the mean values of LGPF, TGPF, LRPF (Leading Vehicle, LV), and TRPF, while insignificantly affecting the mean value of LRPF (Following Vehicle, FV). Nevertheless, a notable discrepancy in the mean value of LRPF (FV) was observed between the warning and guidance groups. Comparative results with lane positions demonstrate that the proposed TPF accurately depicts vehicle lane-changing behaviors. Comparative results with classical risk indicators indicate that TPF more accurately depicts the actual driving risks and variation trends faced by vehicles in different MLC processes. The research findings will be pivotal in improving real-time control and guidance strategies for vehicles in highway merging sections under ICV environments.