A novel collision avoidance strategy for highway overtaking considering the driver's steering intent
Zijun Zhang,Weihe Liang,Han Zhang,Wanzhong Zhao,Chunyan Wang,Heng Huang
DOI: https://doi.org/10.1177/09544070241232137
2024-05-03
Proceedings of the Institution of Mechanical Engineers Part D Journal of Automobile Engineering
Abstract:Proceedings of the Institution of Mechanical Engineers, Part D: Journal of Automobile Engineering, Ahead of Print. Intelligent driving has been prevailing worldwide and is also challenging, which can be complicated by the factors of human drivers. In this paper, a novel collision avoidance strategy is proposed to enhance driving safety in highway overtaking by comprehensively considering the driver's steering intent. First, in order to capture the driver's operational characteristics from the driving data, we formulate the prediction of the driver's steering intent and the ego vehicle's states as a multivariate time series (MTS) forecasting problem, which is then handled by deep learning with a time pattern attention mechanism (DL-Attn). Second, a predictive risk field (PRF) model is proposed to quantify the real-time overtaking risk based on the above prediction results. Then, the overtaking is evaluated via a personalized risk threshold which can be set for a specific driver via experiments. Next, a linear time-varying model predictive control (LTV-MPC) -based assistance controller is designed so as to interfere in the risky overtaking and take over the ego vehicle from the driver to avoid possible collisions. And the feasibility and stability of the closed system are ensured theoretically. Finally, experiments are carried out in three typical cases. The results demonstrate that the proposed strategy can not only effectively improve driving safety for highway overtaking, but also identify safe overtaking to avoid unnecessary interference.
engineering, mechanical,transportation science & technology