Lane Change Intention Recognition for Human Driving Vehicles under Moving Bottleneck on Freeway

Abstract

Accurate lane change intention prediction can assist vehicles in reducing the probability of crashes, which is essential for preserving traffic safety. This paper proposes a data-driven model for manual vehicle lane change intention recognition under freeway moving bottlenecks. Target vehicles are classified by K-means++ algorithm based on their driving behavior. Then, the lane change intention recognition model based on long short-term memory (LSTM) network is built by taking into account the driving behavior characteristics of the target vehicle and surrounding vehicles, the size of surrounding vehicles, and the type of the target vehicle. The Highway Drone (HighD) Dataset collected from German Highways is used for validation. Vehicles from the dataset are categorized as “efficient and experienced”, “safe and sound”, “efficient and reckless” and “safe and cautious”. The results show that the suggested model has an accuracy of 91.63% in recognizing vehicle lane change intention, which is 10.38% better than not taking moving bottlenecks and clustering analysis into consideration. The research can be used to predict lane change intentions of the surrounding manual vehicles under moving bottleneck environment for freeway drivers, and to assist intelligent vehicle driving decision system to ensure driving safety.