利用指数损失因子提高长短期记忆网络轨迹预测精度的方法

兰州理工大学学报 ›› 2023, Vol. 49 ›› Issue (5): 86-92.

• 自动化技术与计算机技术 • 上一篇下一篇

利用指数损失因子提高长短期记忆网络轨迹预测精度的方法

张彤¹, 王志文^*1,2,3, 卢延荣¹, 孙洪涛⁴

1.兰州理工大学电气工程与信息工程学院, 甘肃兰州 730050;
2.兰州理工大学甘肃省工业过程先进控制重点实验室, 甘肃兰州 730050;
3.兰州理工大学电气与控制工程国家级实验教学示范中心, 甘肃兰州 730050;
4.曲阜师范大学工学院, 山东曲阜 273100

收稿日期:2021-12-13 出版日期:2023-10-28 发布日期:2023-11-07
通讯作者: 王志文(1976-),男,甘肃武威人,博士,教授,博导. Email:wzw@lut.edu.cn
基金资助:
国家自然科学基金(61863026),甘肃省高等学校产业支撑引导项目(2019C-05),甘肃省工业过程先进控制重点实验室开放基金(2019KFJJ03),甘肃省重大专项(21ZD4GA028)

Improve the LSTM trajectory prediction accuracy through an exponential weighted loss function

ZHANG Tong¹, WANG Zhi-wen^1,2,3, LU Yan-rong¹, SUN Hong-tao⁴

1. College of Electrical and Information Engineering,Lanzhou Univ. of Tech., Lanzhou 730050, China;
2. Key Laboratory of Gansu Advanced Control for Industrial Processes,Lanzhou Univ. of Tech., Lanzhou 730050, China;
3. National Demonstration Center for Experimental Electrical and Control Engineering Education, Lanzhou Univ. of Tech., Lanzhou 730050, China;
4. College of Engineering, Qufu Normal University, Qufu 273100, China

Received:2021-12-13 Online:2023-10-28 Published:2023-11-07

摘要/Abstract

摘要： 长短期记忆(LSTM)网络在车辆轨迹预测任务中取得了一定的成效.然而,LSTM网络预测时会引入累积误差,导致轨迹预测的精度随预测时间逐渐降低.如何减小LSTM网络在轨迹预测过程中的累积误差是一个关键性问题.为解决此问题,使用随预测时间增长的指数函数对平滑的L1损失函数进行加权来降低累积误差,使模型能更加准确地预测出车辆的未来轨迹.为验证此方法的有效性,以长短期记忆网络编码器解码器模型为基础,在美国交通部下一代仿真计划(NGSIM)采集的US-101和I-80数据集上进行验证.结果表明,与近年来的深度学习方法相比,本文提出的轨迹预测方法0.2 s时刻平均误差从0.332 0 m降低到0.103 2 m,且5 s时刻平均误差由原来的7.716 8 m降为6.624 3 m,总体上比原来降低了14.16%,有效减小了累积误差.

关键词: 自动驾驶, 轨迹预测, LSTM, 编码器-解码模型, 指数加权的损失函数

Abstract: Although a long-short-term memory (LSTM) network has been widely adopted to predict the vehicle trajectory, the iterative nature of LSTM introduces accumulative errors during training, resulting in a gradual decrease in the accuracy of trajectory prediction over time. Therefore, how to reduce the accumulative errors of the LSTM is a critical issue. To solve this problem, an exponential weighted loss function is used in this paper to weight the smoothed L1 loss function to reduce the cumulative error, enabling the model to predict the future trajectory of the vehicle more accurately, which is validated on US-101 and I-80 datasets from Next Generation Simulation (NGSIM). The simulation results show that the test dataset’s average error at 0.2 s and 5 s is reduced from 0.332 0 m and 7.716 8 m to 0.103 2 m and 6.624 3 m, respectively. The average prediction error is reduced by 14.16%, significantly reducing the cumulative error.

Key words: autonomous driving, trajectory prediction, LSTM, an encoder-decoder model, an exponential loss function

中图分类号:

TP3
U461.91

张彤, 王志文, 卢延荣, 孙洪涛. 利用指数损失因子提高长短期记忆网络轨迹预测精度的方法[J]. 兰州理工大学学报, 2023, 49(5): 86-92.

ZHANG Tong, WANG Zhi-wen, LU Yan-rong, SUN Hong-tao. Improve the LSTM trajectory prediction accuracy through an exponential weighted loss function[J]. Journal of Lanzhou University of Technology, 2023, 49(5): 86-92.

参考文献

[1] NILSSON J,SILVLIN J,BRANNSTROM M,et al.If,when,and how to perform lane change maneuvers on highways [J].IEEE Intelligent Transportation Systems Magazine,2016,8(4):68-78.
[2] 张志远,倪国新,徐艳国.轨迹预测技术的现状及发展综述 [J].电子测量技术,2020,43(13):111-116.
[3] JAIN A,SINGH A,KOPPULA H S,et al.Recurrent neural networks for driver activity anticipation via sensory-fusion architecture [C]//IEEE International Conference on Robotics and Automation.Stockholm:IEEE,2016:3118-3125.
[4] LEE N,CHOI W,VERNAZA P,et al.DESIRE:distant future prediction in dynamic scenes with interacting agents[C]//30th IEEE Conference on Computer Vision and Pattern Recognition.Honolulu:IEEE,2017:2165-2174.
[5] GUPTA A,JOHNSON J,LI F F,et al.Socially acceptable trajectories with generative adversarial networks [C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).Salt Lake City:IEEE,2018:2255-2264.
[6] ALAHI A,GOEL K,RAMANATHAN V,et al.Social LSTM:human trajectory prediction in crowded spaces [C]//2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).Seattle:IEEE,2016:961-971.
[7] 曹凯,于善义,于少伟.基于多隐马尔可夫模型的车辆机动行为识别与预测 [J].信息与控制,2014,43(4):506-512.
[8] GENG X L,LIANG H W,YU B,et al.A scenario adaptive driving behavior prediction approach to urban autonomous driving [J].Applied Sciences-Basel,2017,7(4):426-416.
[9] ZIEBART B D,MASS A L,DEY A K,et al.Navigate like a cabbie:probabilistic reasoning from observed context-aware behavior [C]//10th International Conference on UbiquitousComputing.Seoul:Assoc Computing Machinery,2008:322-331.
[10] MORZY M.Mining frequent trajectories of moving objects for location prediction [C]//5th International Conference on Machine Learning and Data Mining in Pattern Recognition.Leipzig:Springer-Verlag Berlin,2007:667-680.
[11] CHO S B.Exploiting machine learning techniques for location recognition and prediction with smartphone logs [J].Neurocomputing,2016,176:98-106.
[12] HOCHREITER S,SCHMIDHUBER J.Long short-term memory [J].Neural Computation,1997,9(8):1735-1780.
[13] ALTCHE F,DE LA FORTELLE A.An LSTM network for highway trajectory prediction [C]//2017 IEEE 20th International Conference on Intelligent Transportation Systems (ITSC).Yokohama:IEEE,2017:353-359.
[14] SCHEEL O,NAGARAJA S,SCHWARZ L,et al.Attention-based lane change prediction [C]//IEEE International Conference on Robotics and Automation ICRA.Montreal:IEEE,2019:8655-8661.
[15] BI H K,FANG Z,MAO T L,et al.Joint prediction for kinematic trajectories in vehicle-pedestrian-mixed scenes [C]//IEEE International Conference on Computer Vision.Seoul:IEEE,2019:10382-10391.
[16] 徐洪敏.LSTM网络在船舶航行轨迹预测中的应用 [J].舰船科学技术,2021,43(8):37-39.
[17] PARK S H,KIM B D,KANG C M,et al.Sequence-to-sequence prediction of vehicle trajectory via LSTM encoder-decoder architecture [C]//2018 IEEE Intelligent Vehicles Symposium (IV).Changshu:IEEE,2018:1772-1678.
[18] WANG C,MA L,LI R,et al.Exploring trajectory prediction through machine learning methods [J].IEEE Access,2019,7(99):101441-101452.
[19] 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.
[20] 王皓昕,李振龙,赵晓华.加权指数损失下长短时记忆网络换道意图识别模型 [J].科学技术与工程,2021,21(1):254-259.
[21] 韩皓,谢天.基于注意力Seq2Seq网络的高速公路交织区车辆变道轨迹预测 [J].中国公路学报,2020,33(6):106-118.
[22] ZHOU Y,LI Z M,XIAO S J,et al.Auto-conditioned recurrent networks for extended complex human motion synthesis [C/OL].[2021-09-15].http://arxiv.org/pdf/1707.05363v3.pdf.
[23] COLYAR J,HALKIAS J.US highway 101 dataset [R].Washington:Federal Highway Administration (FHWA),2007:10-16.
[24] COLYAR J,HALKIAS J.US highway 80 dataset [R].Washington:Federal Highway Administration (FHWA),2006:36-42.
[25] DEO N,TRIVEDI M M.Multi-modal trajectory prediction of surrounding vehicles with maneuver based LSTMs [C] //2018 IEEE Intelligent Vehicles Symposium(IV).Changshu:IEEE,2018:1179-1184.
[26] DEO N,TRIVEDI M M.Convolutional social pooling for vehicle trajectory prediction[C]//IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops.Salt Lake City:IEEE,2018:1549-1557.

利用指数损失因子提高长短期记忆网络轨迹预测精度的方法

Improve the LSTM trajectory prediction accuracy through an exponential weighted loss function

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 15

编辑推荐

Metrics

本文评价

[1]	颜闽秀, 张萍. 保守系统的类混沌吸引子共存分析及图像加密[J]. 兰州理工大学学报, 2023, 49(5): 102-111.
[2]	袁小鹏, 苗思蕊, 马蓉, 冯涛, 方君丽. 基于区块链的社会化考试成绩安全管理系统[J]. 兰州理工大学学报, 2023, 49(4): 88-94.
[3]	李洋, 万洪妤, 朱建明, 王秀利, 王友卫. 基于数字承诺的去中心化区块链交易方案[J]. 兰州理工大学学报, 2023, 49(4): 95-101.
[4]	陈勇群, 陈玉成, 胡华淼, 戴佳浩, 康潆允, 王巍. 基于生成对抗网络的IPv6地址存活性预测[J]. 兰州理工大学学报, 2023, 49(4): 102-107.
[5]	王星, 晏榕璟. 深度可分卷积结合多通道注意力的垃圾图像快速分类模型[J]. 兰州理工大学学报, 2023, 49(3): 88-93.
[6]	谢楠, 马振国, 唐兵, 黄煜铭, 张柯琪, 曹丹怡. 继电保护设备剩余寿命预测的智能算法研究[J]. 兰州理工大学学报, 2023, 49(2): 83-87.
[7]	苏昀暄, 涂正, 王绪安, 林川. 一种基于CKKS同态加密与神经网络的安全人脸识别方案[J]. 兰州理工大学学报, 2023, 49(2): 103-109.
[8]	王燕, 赵妮妮, 范林. 基于图神经网络特征交叉的协同过滤算法[J]. 兰州理工大学学报, 2023, 49(1): 94-102.
[9]	赵宏, 郑厚泽, 郭岚. 基于词-标签概率的多标签文本分类研究[J]. 兰州理工大学学报, 2023, 49(1): 103-109.
[10]	赵荣珍, 赵楠. 结合扰动集成RBF的故障识别方法[J]. 兰州理工大学学报, 2022, 48(6): 40-45.
[11]	卢鹏丽, 陈云天. 基于网络拓扑和多种生物信息融合的关键蛋白质识别算法[J]. 兰州理工大学学报, 2022, 48(6): 96-103.
[12]	陈朴, 方景芳, 刘军, 陈家伟. 汽车轮渡智能装载优化方案研究[J]. 兰州理工大学学报, 2022, 48(6): 104-109.
[13]	冯笑, 杨雅婷, 董瑞, 艾孜麦提·艾尼瓦尔, 马博. 基于回译和集成学习的维汉神经机器翻译方法[J]. 兰州理工大学学报, 2022, 48(5): 99-106.
[14]	郭润兰, 史方青, 范雅琼, 何智. 基于卷积神经网络的壁面清洗机器人障碍物检测识别算法[J]. 兰州理工大学学报, 2022, 48(4): 83-89.
[15]	卢鹏丽, 许星舟. 基于精准k核的复杂网络节点重要性评估方法[J]. 兰州理工大学学报, 2022, 48(4): 90-98.