基于注意力与深度交互的周车多模态行为轨迹预测

Multi-mode behavior trajectory prediction of surrounding vehicle based on attention and depth interaction

设计了一种车辆深度交互编码并结合基于注意力机制的解码器模型,该模型同时输出车辆多模态行为预测结果和未来轨迹预测分布。使用公开的NGSIM US-101和I-80数据集评估所提出的模型,并且对模型多模态行为机动预测进行了定性分析。结果表明:该模型具有较好的均方根误差值(RMSE),在提升了计算效率的基础上获得了更高的轨迹预测精度。

A vehicle deep-interaction coding model combined with a decoder based on the attentionmechanism to solve this problem was presented in this work.The model′s multi-modal behavior predictions and trajectory predictions were output.The proposed model is evaluated by the public NGSIM US-101 and I-80 data sets.The results show that the model has a better root mean square error and achieves higher trajectory prediction accuracy while improving computational efficiency.This paper also shows a qualitative analysis of the prediction of multi-modal behavior maneuvering.