智能驾驶车辆轨迹预测方法综述

A Survey of Vehicle Trajectory Prediction Methods for Intelligent Driving

针对目前车辆轨迹预测难点,对车辆轨迹预测方法的分类和研究现状进行综述。根据模型实现预测时域的不同,将现有算法分为短时域和长时域的车辆轨迹预测方法;介绍短时域的基于物理模型和传统机器学习预测方法的基本概念及研究现状,总结对比长时域的基于深度学习、神经网络和基于车辆驾驶行为意图识别的预测方法。分析结果表明:长时域方法能够解决车辆轨迹预测难点问题,保证智能车辆高效、安全驾驶。

In view of the difficulties of vehicle trajectory prediction,the classification and research status of vehicle trajectory prediction methods are reviewed.According to the different prediction time domain of model implementation,the existing algorithms are divided into short time domain and long time domain vehicle trajectory prediction methods;the basic concepts and research status of short time domain prediction methods based on physical model and traditional machine learning are introduced,and the long time domain prediction methods based on deep learning,neural network and vehicle driving behavior intention recognition are summarized and compared.The analysis results show that the long time domain method can solve the difficult problem of vehicle trajectory prediction and ensure the efficient and safe driving of intelligent vehicles.