基于BiGRU的多模态驾驶行为及轨迹预测

删除或更新信息，请邮件至freekaoyan#163.com(#换成@)

本站小编 Free考研考试/2024-01-16

-->

包智鹏,支永帅,张素民,何睿.基于BiGRU的多模态驾驶行为及轨迹预测[J].,2021,61(3):246-254

基于BiGRU的多模态驾驶行为及轨迹预测

BiGRU based multi modal maneuvers and trajectory prediction

DOI：10.7511/dllgxb202103004

中文关键词:车辆工程驾驶行为预测轨迹预测 BiGRU 车辆交互

英文关键词:automotive engineering maneuver prediction trajectory prediction BiGRU vehicle interaction

基金项目:国家重点研发计划项目(2016YFB0100904)；国家自然科学基金资助项目(U1564211)；吉林大学研究生创新基金资助项目(101832020CX138).

作者	单位
包智鹏,支永帅,张素民,何睿

摘要点击次数:272

全文下载次数:360

中文摘要:

在复杂交通环境中行驶的智能汽车需要预测未来周围车辆的动向，为了提升智能汽车快速且准确预测周围车辆驾驶行为及轨迹的能力，设计了一种基于BiGRU的多模态驾驶行为及轨迹预测模型．模型由BiGRU编码器、交互卷积池化层和GRU解码器组成，能够预测未来5 s车辆多模态驾驶行为的概率和多模态驾驶行为对应的轨迹分布．试验结果表明，相较于其他基于深度学习的模型，该模型在预测长时域轨迹时的RMSE损失和NLL损失更低，具备更高的准确率．

英文摘要:

Forecasting the motion of surrounding vehicles is a critical ability for an intelligent vehicle deployed in the complex traffic. In order to improve the ability of intelligent vehicles to predict the maneuvers and trajectory of surrounding vehicles quickly and accurately, a BiGRU based multi modal maneuvers and trajectory prediction model is proposed. The model consists of a BiGRU encoder, a convolution social pooling and a GRU decoder, which can predict the probability of multi modal maneuvers and the corresponding trajectory distribution of multi modal maneuvers in the future 5 s. Experimental results indicate that compared with other deep learning based models, this model has lower RMSE loss and NLL loss in the prediction of long term trajectory, and has higher accuracy.

查看全文查看/发表评论下载PDF阅读器

关闭

基于BiGRU的多模态驾驶行为及轨迹预测

本站小编 Free考研考试/2024-01-16

相关话题/

领限时大额优惠券,享本站正版考研考试资料!