基于变分自动编码器的车辆轨迹预测研究

针对轨迹预测中车辆与周边车辆、道路几何之间交互关系建模不充分，以及车辆轨迹多模态建模不完整等一系列问题，提出了一种基于变分自动编码器的车辆轨迹预测方法。首先，通过长短时记忆网络从原始数据中提取轨迹数据与车道信息的语义特征；其次，引入多头注意力机制，采用两个单独的注意力模块分别建立车辆与车辆交互模型及车辆与道路交互模型，能够更好地反映周边车辆与道路几何对车辆轨迹的交互影响，得到丰富的场景上下文信息；接着利用变分自动编码器对车辆轨迹多模态建模，捕捉轨迹预测的随机性质以生成合理的未来轨迹分布；最后从分布中多次重复采样以生成多条可能的未来轨迹。通过搭建实验平台和使用Argoverse自然驾驶数据集进行测试，改进后的预测方法在平均位移误差和最终位移误差指标下的数值分别为1.03和1.51,预测精度上相较于其他3种预测方法，分别提升了45%、46%、32%;实验结果表明：预测方法可以有效地改善车辆与周边车辆、道路几何之间交互关系建模不充分，以及车辆轨迹多模态建模不完整等问题，预测精度提高，总体预测性能良好。

Research on Vehicle Trajectory Prediction Based on Variational Automatic Encoder

A vehicle trajectory prediction method based on variational automatic coder was proposed to address a series ofproblems such as inadequate modeling of the interaction between vehicles and surrounding vehicles and road geometry intrajectory prediction and incomplete multimodal modeling of vehicle trajectories. Firstly the semantic features oftrajectory data and lane information were extracted from the original data through long and short term memory networks.Secondly a multi-headed attention mechanism was introduced and two separate attention modules were used to establishthe vehicle-vehicle interaction model and vehicle-road interaction model respectively which can better reflect theinteraction effects of surrounding vehicles and road geometry on vehicle trajectories and obtain rich scene contextinformation. Next the vehicle trajectory was modelled multimodally by using the variational automatic coder to capture therandom nature of trajectory prediction to generate a reasonable future trajectory distribution. Finally the sampling wasrepeated several times from the distribution to generate multiple possible future trajectories. By building the experimentalplatform and testing with the Argoverse naturalistic driving dataset the improved prediction method yielded values of 1. 03and 1. 51 under the average displacement error and final displacement error indicators respectively and the prediction accuracy has been improved by 45% 46% and 32% compared with the other three prediction methods. Theexperimental results showed that the prediction method can effectively solve the problems of inadequate modeling of theinteraction between vehicles and surrounding vehicles and road geometry and incomplete multi-modal modeling of vehicletrajectories. The prediction accuracy has been improved and the overall prediction performance is good.