Vehicle trajectory prediction based on LSTM network

Abstract

In a complex traffic environment, predicting the trajectory of surrounding vehicles in the driver's line of sight can greatly reduce the possibility of various traffic accidents and play an auxiliary role in the driver's decision making. The motion of predicted vehicles is constrained by the traffic environment, that is, the motion of adjacent vehicles and the relative spatial positions between vehicles. This paper mainly studies the behavior prediction of vehicles on the expressway. Based on the social convolutional pooling LSTM network (CS-LSTM), a CS-LSTM network with an attention mechanism is proposed, which assigns different weights to the fusion features and improves the accuracy of the trajectory prediction of surrounding vehicles. This article evaluates the model on a publicly available NGSIM dataset. The results show that the proposed algorithm is more accurate than other algorithms in predicting the future trajectory of vehicles.