Please wait a minute...
浙江大学学报(工学版)
浙江大学学报(工学版)  2020, Vol. 54 Issue (6): 1156-1163    DOI: 10.3785/j.issn.1008-973X.2020.06.012
计算机技术     
基于注意力机制的车辆运动轨迹预测
刘创(),梁军*()
浙江大学 控制科学与工程学院，浙江 杭州 310058
Vehicle motion trajectory prediction based on attention mechanism
Chuang LIU(),Jun LIANG*()
College of Control Science and Engineering, Zhejiang University, Hangzhou 310058, China
 全文: PDF(972 KB)   HTML
摘要：

基于经典的Convolutional Social LSTM轨迹预测算法，提出一种全新的采用注意力机制的车辆运动轨迹预测算法. 引入横向注意力机制对邻居车辆赋予不同的权重，将车辆历史轨迹经由LSTM得到的特征作为全局特征，通过卷积池化提取轨迹特征作为局部特征，将两者融合作为整体邻居特征信息，用于轨迹预测. 对用于传统轨迹预测的Encoder-Decoder框架进行改进，引入关于历史位置的纵向注意力机制，使得预测的每一时刻都能使用与当前时刻最相关的历史信息. 在NGSIM提供的US101和I80数据集进行验证，结果表明：提出的轨迹预测算法相比其他算法能更精确地预测车辆未来轨迹.
关键词： 自动驾驶;  轨迹预测;  注意力机制;  长短期记忆（LSTM）    
Abstract:

A new vehicle motion trajectory prediction algorithm was proposed by using the attention mechanism based on the classic convolutional social long-short term memory (LSTM) trajectory prediction algorithm. Firstly, the lateral attention mechanism was introduced to assign different weights to neighboring vehicles. The features obtained from the historical trajectory of the vehicle via LSTM were taken as global features, and the trajectory features were extracted as local features through convolution pooling. The two features were fused as the overall neighbor feature information for trajectory prediction. Secondly, the Encoder-Decoder framework of traditional trajectory prediction was improved, and a vertical attention mechanism on historical position was introduced, so that each moment of prediction could use the historical information, which was most relevant to the current moment. The improved model was verified on the US101 and I80 datasets provided by NGSIM, and the results show that the proposed trajectory prediction algorithm can obtain more accurate future trajectories than other algorithms.
Key words: self-driving    trajectory prediction    attention mechanism    long-short term memory (LSTM)
收稿日期: 2019-01-22 出版日期: 2020-07-06
CLC:  TP 183  
基金资助: 国家自然科学基金资助项目（U1664264，2019YFB1600500）
通讯作者: 梁军     E-mail: 21632012@zju.edu.cn;jliang@zju.edu.cn
作者简介: 刘创（1995—），男，硕士生，从事计算机视觉研究. orcid.org/0000-0003-2259-9440. E-mail： 21632012@zju.edu.cn
服务　　
把本文推荐给朋友
加入引用管理器
E-mail Alert
作者相关文章　　
刘创
梁军

引用本文:

刘创,梁军. 基于注意力机制的车辆运动轨迹预测[J]. 浙江大学学报(工学版), 2020, 54(6): 1156-1163.

Chuang LIU,Jun LIANG. Vehicle motion trajectory prediction based on attention mechanism. Journal of ZheJiang University (Engineering Science), 2020, 54(6): 1156-1163.

链接本文:

http://www.zjujournals.com/eng/CN/10.3785/j.issn.1008-973X.2020.06.012    或    http://www.zjujournals.com/eng/CN/Y2020/V54/I6/1156

图 1  社会池化过程示意图
图 2  社会池化网络框架
图 3  卷积社会池化网络框架
图 4  基于注意力机制的轨迹预测模型
预测算法 1 s 2 s 3 s 4 s 5 s $\delta $
KF 1.936 3.860 5.779 7.704 9.648 5.013
LSTM 0.698 1.691 2.981 4.567 6.465 2.741
Fc_pool 0.673 1.518 2.584 3.910 5.524 2.382
Conv_pool 0.661 1.472 2.509 3.821 5.406 2.324
Nbrs_conv 0.635 1.448 2.470 3.751 5.306 2.280
Hist_Nbrs_conv 0.629 1.432 2.454 3.730 5.265 2.263
表 1  各模型轨迹预测均方误差对比
图 5  各模型轨迹偏差对比
图 6  真实数据集上模型轨迹预测结果对比
图 7  车辆注意力权重分布图
图 8  不同时刻下的历史轨迹的权重分布
1 HELBING D, MOLNAR P Social force model for pedestrian dynamics[J]. Physical Review E, 1995, 51 (5): 4282
doi: 10.1103/PhysRevE.51.4282
2 PELLEGRINI S, ESS A, SCHINDLER K, et al. You'll never walk alone: modeling social behavior for multi-target tracking [C] // 2009 IEEE 12th International Conference on Computer Vision. Kyoto: IEEE, 2009: 261-268.
3 HOCHREITER S, SCHMIDHUBER J Long short-term memory[J]. Neural Computation, 1997, 9 (8): 1735- 1780
doi: 10.1162/neco.1997.9.8.1735
4 ALAHI A, GOEL K, RAMANATHAN V, et al. Social LSTM: Human trajectory prediction in crowded spaces [C] // Proceedings of the IEEE Conference On Computer Vision and Pattern Recognition. Las Vegas: IEEE 2016: 961-971.
5 XU K, QIN Z, WANG G, et al. Collision-Free LSTM for human trajectory prediction [C] // International Conference on Multimedia Modeling. Bangkok: Springer, Cham, 2018: 106-116.
6 DEO N, TRIVEDI M M. Convolutional social pooling for vehicle trajectory prediction [C] // Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops. Munich: IEEE, 2018: 1468-1476.
7 GUPTA A, JOHNSON J, FEI-FEI L, et al. Social GAN: socially acceptable trajectories with generative adversarial networks [C] // Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Salt Lake City: IEEE, 2018: 2255-2264.
8 MANH H, ALAGHBAND G. Scene-LSTM: a model for human trajectory prediction [J]. arXiv Preprint, arXiv: 1808.04018, 2018.
9 XUE H, HUYNH D Q, REYNOLDS M. SS-LSTM: a hierarchical LSTM model for pedestrian trajectory prediction [C] // 2018 IEEE Winter Conference on Applications of Computer Vision (WACV). Lake Tahoe: IEEE, 2018: 1186-1194.
10 FERNANDO T, DENMAN S, SRIDHARAN S, et al Soft+ hardwired attention: an LSTM framework for human trajectory prediction and abnormal event detection[J]. Neural Networks, 2018, 108: 466- 478
doi: 10.1016/j.neunet.2018.09.002
11 GRAVES A. Generating sequences with recurrent neural networks[J]. arXiv Preprint, arXiv: 1308.0850, 2013.
12 COLYAR J, HALKIAS J. US highway 101 dataset [R]. Washington: Federal Highway Administration (FHWA), 2007.
[1] 雍子叶,郭继昌,李重仪. 融入注意力机制的弱监督水下图像增强算法[J]. 浙江大学学报(工学版), 2021, 55(3): 555-562.
[2] 张岩,郭斌,王倩茹,张靖,於志文. SeqRec：基于长期偏好和即时兴趣的序列推荐模型[J]. 浙江大学学报(工学版), 2020, 54(6): 1177-1184.
[3] 赵小虎,尹良飞,赵成龙. 基于全局?局部特征和自适应注意力机制的图像语义描述算法[J]. 浙江大学学报(工学版), 2020, 54(1): 126-134.
[4] 董月,冯华君,徐之海,陈跃庭,李奇. Attention Res-Unet: 一种高效阴影检测算法[J]. 浙江大学学报(工学版), 2019, 53(2): 373-381.
[5] 郭宝震, 左万利, 王英. 采用词向量注意力机制的双路卷积神经网络句子分类模型[J]. 浙江大学学报(工学版), 2018, 52(9): 1729-1737.