A Following Model Considering Multiple Vehicles from the Driver's Front and Rear Perspectives
Weiwei Qi,Wenyi Wang,Chuanyun Fu
DOI: https://doi.org/10.1016/j.physa.2024.130174
2024-01-01
Abstract:In the current traffic environment dominated by manual driving, existing models of drivers' rearview perceptions are inadequate. Existing car-following models that incorporate rear-view information are primarily focused on the Internet of Vehicles (IoV) and automated driving environments. However, they fail to realistically reflect the visual processing mechanisms of human drivers, limiting their effectiveness in realistic traffic scenarios. Therefore, we propose a new following model, the multi-vehicle influence from front and rear perspectives (MVFR), that considers the influence of multiple vehicles. The MVFR model combines information from both front and rear vehicles, integrating views from the front, side front and rear. It provides an indepth analysis of the effects of relative state differences between a vehicle and its surrounding vehicles on speed, including the effects of perspectives in both the lateral and longitudinal directions. Linear stability analysis and numerical simulation demonstrate that considering the perspectives of rear-following vehicles and lateral offset angles can improve traffic flow stability to a certain extent. Furthermore, properly considering the lateral offset distance and the number of vehicles ahead also positively affects traffic flow stability. This study reveals that observing following vehicles and considering information from multiple front vehicles enhances system stability, especially when there is no or minimal lateral offset. In contrast, focusing on fewer front vehicles is more effective for traffic flow stability when there is a large lateral offset. Experimental results using the CKQ4up dataset show that the MVFR model achieves higher accuracy than the conventional FVD model, the front-view-only improved FVD model (MFVD-RV), and the MFRHVAD-RV and MFRHVAD-AV models. Compared with models relying solely on front-view or non-visual perception, the MVFR model demonstrates a better fit, validating the advantages of this full-view perception model in manual driving environments. This innovation addresses the shortcomings of existing research, thereby enhancing the reliability of models under manual driving conditions on highways.