C.H. Sun, Y. Sun

Pages: 55-64

Abstract
The traditional lane change trajectory prediction method has some problems such as large deviation of actual estimation and long prediction time. This paper proposes a lane change trajectory prediction method in the network of vehicles environment. Firstly, the hidden Markov model is used to classify the vehicle behavior in networked vehicle environment into three types: left lane change, right lane change and straight lane change. Secondly, according to the vehicle behavior, the lateral displacement is taken as the index to judge the safety state of the vehicle, and the lane-changing intention of the vehicle is analyzed by using the gradient lifting decision tree. Finally, the vehicle state vector is mapped to the social pool according to the result of lane change intention, combined with the vehicle speed and Angle, and the road network information, traffic control information, road traffic flow information, traffic control state information in the Internet of vehicles and real-time traffic environment information are used as lane change trajectory prediction data. Combined with vehicle state vector, lane change trajectory prediction is realized by multi-layer perceptron. The experimental results show that the proposed method has a high degree of fitting with the actual trajectory under different time domain conditions, and the root mean square error of prediction is stable within 0.64, and the prediction time is short.
Keywords: internet of vehicles environment; lane-changing trajectory; hidden markov model; gradient lifting decision tree; training mechanism; intention analysis