基于边缘计算的交叉路口无人驾驶车辆通行轨迹预测算法

doi:10.3969/j.issn.1674-8484.2021.02.003

摘要/Abstract

摘要：

为给无人驾驶车辆精确制定城市路口通行轨迹提供先验信息,提出了在边缘计算平台中基于驾驶员意图分类和Bezier曲线相结合的轨迹预测算法。分析2个路口230辆车的真实通行数据,提出了基于支持向量机的驾驶员意图识别算法,预测路口车辆直行、左转及右转的概率;用基于Bezier曲线和代价函数相结合的通行轨迹预测方法,预测其路口通行轨迹。结果表明：经过与采集的120辆车的实际数据对比,驾驶员意图分类算法准确度在92.5%以上,车辆预测轨迹与真实轨迹间最大偏差范围在22.3~57.9 cm之间,所有车辆预测轨迹与真实轨迹间平均偏差为21.4 cm。因此,本方法能够满足城市路口车辆通行轨迹预测需求。

关键词: 无人驾驶车辆, 城市路口通行, 轨迹预测, 驾驶员意图识别, 边缘计算

Abstract:

A trajectory prediction algorithm was proposed based on the combination of driver intent classification and Bezier curve in the edge computing platform to provide prior information for unmanned vehicles to accurately formulate urban intersection trajectories. Proposed a support vector machine-based driver intention recognition algorithm to predict the probability of vehicles going straight, turning left or turning right at the intersection after analyzing the actual traffic data of 230 vehicles at two intersections. And proposed a traffic trajectory prediction method based on the combination of Bezier curve and cost function to predict the traffic trajectory of the intersection. The results show that the accuracy of the driver’s intention classification algorithm is above 92.5% comparing with the actual data collected from 120 vehicles. The maximum deviation range between the predicted trajectory of the vehicle and the true trajectory is 0.223~0.579 m. The average deviation between the predicted trajectory of all vehicles and the true trajectory is 0.214 m. Therefore, this method meets the needs of vehicle trajectory prediction at urban intersections.

Key words: unmanned vehicles, urban intersection transiting, trajectory prediction, driver intention recognition, edge computing

中图分类号:

U467.1+4

郝璐璐, 谢辉, 宋康, 闫龙. 基于边缘计算的交叉路口无人驾驶车辆通行轨迹预测算法[J]. 汽车安全与节能学报, 2021, 12(2): 163-172.

HAO Lulu, XIE Hui, SONG Kang, YAN Long. Trajectory prediction algorithm of unmanned vehicles at urban intersection based on edge computing[J]. Journal of Automotive Safety and Energy, 2021, 12(2): 163-172.

图/表 22

图1 智慧道路路线图红色数字表示路口编号;蓝色路线为高精度地图,包含路网数据、车道线数据、道路边沿数据、道拓扑结构数据等数据。

图2 激光雷达感知效果图蓝色方块一机动车;绿色方块一非机动车。

硬件设备	性能指标	输出信息
激光雷达	水平视场角：360°;垂直视场角：30°; 最大测距：150 m;精度：±2 cm。	种类、ID、长宽高、速度、航向角、经纬度时间戳、相对角度、相对距离。
Xavier	内存：16 GB;CPU：8个;架构：ARM;接口：以太网/CAN、总线/串口/USB等。	处理激光雷达点云数据。

表1 传感器参数表

图3 路侧通讯架构图

图4 交叉路口ROI图

图5 路口采集车辆轨迹簇

图6 路口车辆行驶示意图

图7 路口车辆数据分布图

图8 驾驶员意图分类模块

图9 驾驶员意图识别概率图

图10 拟合曲线选择

图11 目标车辆起终点示意图

图12 距离和航向角的拟合等高图

图13 目标车辆控制点示意图

图14 9号路口拟合车辆轨迹簇

图15 9号路口预测车辆轨迹簇

输入特征向量		准确率/%
个数	种类	准确率/%
1	θ	58.7
1	Δα	40.7
1	v	20.8
1	Δd	22.8
2	(x, y)	69.6

表2 单个特征向量测试结果

输入特征向量		准确率/%
个数	种类	准确率/%
2	θ、Δα	76.6
3	(x, y)、θ	84.5
4	θ、Δα、v、Δd	83.4
4	(x, y)、θ、Δα	85.0
5	(x, y)、θ、Δα、v	91.2
5	(x, y)、θ、Δα、Δd	92.6
6	(x, y)、θ、Δα、Δd、v	92.1

表3 组合特征向量测试结果

路口编号	车辆总数	正确数量	错误数量
5	15	14	1
7	25	23	2
8	35	32	3
9	45	42	3
总计	120	111	9

表4 路口驾驶员意图识别测试结果

图16 调用SVM运行时间

图17 预测轨迹与真实轨迹对比

图18 路口车辆真实轨迹与预测轨迹对比

参考文献 19

[1]	徐彪, 张放, 王建强, 等. 连续交叉路口通行辅助系统[J]. 汽车工程, 2016,38(11):1344-1350.
	XU Biao, ZHANG Fang, Wang Jianqiang, et al. Continuous intersection traffic assistance system[J]. Automotive Engineering, 2016,38(11):1344-1350. (in Chinese)
[2]	邹渊, 张涛, 张旭东, 等. 考虑车流影响的网联车辆节能驾驶研究[J]. 汽车工程, 2020(10):1320-1326.
	ZOU Yuan, ZHANG Tao, ZHAGN Xudong, et al. Research on energy-saving driving of connected vehicles considering the influence of traffic flow[J]. Automotive Engineering, 2020(10):1320-1326. (in Chinese)
[3]	XU Ting, JIANG Ruisen, WEN Changlei, et al. A hybrid model for lane change prediction with V2X-based driver assistance[J]. Physica A: Statal Mech Appl, 2019,534:2-14.
[4]	Schubert R, Richter E, Wanielik G. Comparison and evaluation of advanced motion models for vehicle tracking[C]// 2008 11th Int’l Conf Info Fusion IEEE, Cologne, Germany, 2008: 1-6.
[5]	Brannstrom M, Coelingh E, Sjoberg J. Model-based threat assessment for avoiding arbitrary vehicale collisions[J]. IEEE Trans Intell Transp Syst, 2010,11(3):658-669. doi: 10.1109/TITS.2010.2048314 URL
[6]	Houenou A, Bonnifait P, Cherfaoui V, et al. Vehicle trajectory prediction based on motion model and maneuver recognition[C]// 2013 IEEE/RSJ Int’l Conf Intell Robo Syst, IEEE, Tokyo, Japan, 2013: 4363-4369.
[7]	XIAO Wei, ZHANG Lijun, MENG Dejian. Vehicle trajectory prediction based on motion model and maneuver model fusion with interactive multiple models[J]. SAE Int’l J Adv Current Practices in Mobility, 2020,2:3060-3071.
[8]	FENG Xidong, CEN Zhepeng, HU Jianming, et al. Vehicle trajectory prediction using intention-based conditional variational autoencoder[C]// 2019 IEEE Intell Transp Syst Conf (ITSC), IEEE, Auckland, New Zealand, 2019: 3514-3519.
[9]	张星洲, 鲁思迪, 施巍松. 边缘智能中的协同计算技术研究[J]. 人工智能, 2019(5):55-67.
	ZHANG Xingzhou, LU Sidi, SHI Weisong. Research on collaborative computing technology in edge intelligence[J]. Arti Intell, 2019(5):55-67. (in Chinese)
[10]	WU Gaoxiang, MIAO Yiming, ZHANG Yu, et al. Energy efficient for UAV-enabled mobile edge computing networks: Intelligent task prediction and offloading[J]. Computer Comm, 2020,150:556-562. doi: 10.1016/j.comcom.2019.11.037 URL
[11]	Bekkouche O, Taleb T, Bagaa M, et al. Edge cloud resource-aware flight planning for unmanned aerial vehicles[C]// 2019 IEEE Wireless Comm Networking Conf (WCNC), IEEE, Marrakesh, Morocco, 2019: 1-7.
[12]	Tomar R S, Verma S, Tomar G S. SVM based trajectory predictions of lane changing vehicles[C]// Int’l Conf Compu Intel Commu Networks IEEE, Gwalior, India, 2011: 716-721.
[13]	冷雨泉, 张会文, 张伟. 机器学习入门到实践[M]. 北京: 清华大学出版社, 2019: 77-86.
	LENG Yuquan, ZHANG Huiwen, ZHANG Wei. Introduction to Practice of Machine Learning [M]. Beijing: Tsinghua University Press, 2019: 77-86. (in Chinese)
[14]	Lam H T, Bouillet E. Flexible sliding windows for kernel regression-based bus arrival time prediction[C]// Euro Conf Mach Learning Princi Pract of Knowledge Discovery in Databases,Springer Int’l Publishing, Porto, Portugal, 2015: 68-84.
[15]	ZHOU Wei, WANG Xiaoping, SUN Haoshui, et al. Trajectory optimization of quadrotor UAV based on Bessel curve[C]// 2019 14th IEEE Int’l Conf Electr Meas Instru (ICEMI), IEEE, Changsha, China, 2019: 1434-1440.
[16]	YUE Ming, WU Xiangmin, GUO Lie, et al. Quintic polynomial-based obstacle avoidance trajectory planning and tracking control framework for tractor-trailer system[J]. Int’l J Control Automa Syst, 2019,17(3):2634-2646.
[17]	DONG Jiajia, WANG Taiyong, ZHANG Yongbin, et al. Applications of improved B-spline curves to 6R robot trajectory optimization[J]. Chin Mech Engi, 2018,29(2):193-200.
[18]	姚连璧, 钱瑾斐. 基于移动最小二乘法的轨迹拟合切线方位角计算[J]. 同济大学学报:自然科版, 2018,46(11):121-125.
	YAO Lianbi, QIAN Jinfei. Azimuth angle calculation of trajectory fitting tangent based on moving least squares method[J]. J Tongji Univ:Nat Sci Ed, 2018,46(11):121-125. (in Chinese)
[19]	Devaurs D, Siméon T, Cortés J. Optimal path planning in complex cost spaces with sampling-based algorithms[J]. IEEE Trans Automation Sci Engi, 2016,13(2):415-424.