基于主动优化的无人驾驶客车实时性运动规划算法

doi:10.3969/j.issn.1674-8484.2020.04.007

摘要/Abstract

摘要：

为提高自动驾驶客车运动规划的实时性、舒适性和安全性,提出了基于主动优化的实时运动规划算法。以车道中心线为参考线,建立Station-Lateral坐标系并综合考虑车辆轨迹平顺性等指标,构建代价函数;限定采样空间保证实时性,产生一组较优的横向偏移和纵向速度组合作为基础轨迹。使用模型预测轨迹生成方法,主动地优化轨迹质量。进行了虚拟仿真测试和实车试验。结果表明：该算法在车载低功耗嵌入式计算平台的计算时间平均为48.3 ms,实时性能够能满足城市道路下典型路况的需求。相比于单一的采样方法,算法的轨迹曲率变化率标准差平均降低18.35%。因而,该算法可使自动驾驶客车具备较好的舒适性和安全性。

关键词: 自动驾驶, 自动驾驶客车, 运动规划, 轨迹生成, 舒适性和安全性

Abstract:

A real-time motion planning algorithm based on initiative optimization was developed for the motion planning of autonomous bus to achieve its real-time, comfort and safety. A Station-Lateral coordinate system was established, regarding the lane center line as the reference line. Cost functions were proposed considering indicators such as comfort. The sampling space was limited to ensure real-time performance. A better combination of lateral offset and longitudinal velocity was generated as the basic trajectory. The kinematics model of autonomous bus was used to generate trajectory, which optimized the trajectory quality initiatively. The virtual simulation test and the real vehicle test were implemented. The results show that the algorithm calculation time is 48.3 ms on average, on the low-power embedded computing platform of the vehicle; the real-time performance meets the typical urban road conditions. the standard deviation of trajectory curvature change rate by using the algorithm is reduced by 18.35% on average compared with that by using single sampling method. Therefore, it makes autonomous bus being better comfort and safety.

Key words: autonomous driving, autonomous buses, motion planning, trajectory generation, comfort and safety

中图分类号:

U469.13

周扬, 谢辉, 肖蓬勃, 刘昊, 修国涛. 基于主动优化的无人驾驶客车实时性运动规划算法[J]. 汽车安全与节能学报, 2020, 11(4): 476-486.

ZHOU Yang, XIE Hui, XIAO Pengbo, LIU Hao, XIU Guotao. Real-time motion planning algorithm for autonomous bus based on initiative optimization[J]. Journal of Automotive Safety and Energy, 2020, 11(4): 476-486.

图/表 20

图1 基于采样和模型优化的规划算法架构图

图2 道路和环境模型

图3 S-L坐标系下坐标点的表示

图4 高精地图航路点

图5 S-L坐标系下的横向轨迹采样

图6 纵向速度采样

图7 车辆的圆形近似方法

图8 车辆运动学自行车模型

图9 Newton法优化轨迹的求解过程

迭代次数	κ_h / (°)	κ_t / (°)	s_term / m
1	0.00	0.00	35.00
2	-1.12	-1.88	32.87
3	-1.20	2.09	31.46

表1 动作参数迭代过程

图10 采样方法并行计算过程

图11 优化方法查找表轨迹

图12 软件在环联合仿真平台结构图

长×宽×高	6.605 m ×2.320 m ×2.287 m
整备质量	5.800 t
前悬长度	1.000 m
后悬长度	1.205 m
轴距	4.400 m
最高车速	69 km?h^-1
驱动电机额定功率	50 kW
驱动电机峰值扭矩	750 Nm
轮胎参数	215/75R 17.5

表2 自动驾驶公交车辆参数

图13 结构化道路运行实验轨迹

图14 结构化道路运行过程

图15 瞬时规划轨迹对比

计算平台	Nvidia Jetson Xavier
CPU参数	ARM Carmel 2.0GHz
实现语言	C++
规划距离	50 m
采样横向轨迹数量	5
采样纵向速度数量	8
模型优化最大迭代次数	5

表3 算法计算参数

采样并行计算	优化查找表预构建	t / ms
未使用	未使用	97.0
使用	未使用	69.5
未使用	使用	76.4
使用	使用	48.3

表4 实时性方法计算时间对比

图16 算法计算时间统计

参考文献 15

[1]	李克强, 戴一凡, 李升波, 等. 智能网联汽车(ICV)技术的发展现状及趋势[J]. 汽车安全与节能学报, 2017,8(1):1-14.
	LI Keqiang, DAI Yifan, LI Shengbo, et al. State-of-the-art and technical trends of intelligent and connected vehicles[J]. J Automotive Safety Energy, 2017,8(1):1-14. (in Chinese)
[2]	Shaikh E A, Dhale A. AGV path planning and obstacle avoidance using Dijkstra's algorithm[J]. Int’l J Appl Innov Engineering & Management, 2013,2(6):77-83.
[3]	Hart P E, Nilsson N J, Raphael B. A formal basis for the heuristic determination of minimum cost paths[J]. IEEE Trans Syst Sci and Cybernetics, 1968,4(2):100-107.
[4]	Koenig S, Likhachev M, Furcy D. Lifelong planning A[J]. Arti Intell*, 2004,155(1-2):93-146.
[5]	Stentz A. Intelligent Unmanned Ground Vehicles[M]. Boston: Springer, 1997: 203-220.
[6]	LaValle S M, Kuffner J J. Rapidly-exploring random trees: Progress and prospects[J]. Algor Comput Robot: New Directions, 2001,2001(5):293-308.
[7]	Kuffner J J, LaValle S M. RRT-connect: An efficient approach to single-query path planning[C]// IEEE Int’l Conf Robot Automation. IEEE, 2000,2:995-1001.
[8]	Kondak K, Hommel G. Computation of time optimal movements for autonomous parking of non-holonomic mobile platforms[C]// IEEE Int’l Conf Robotics and Automation. IEEE, 2001,3:2698-2703.
[9]	Madås D, Nosratinia M, Keshavarz M, et al. On path planning methods for automotive collision avoidance[C]// 2013 IEEE Intelligent Vehicles Symposium (IV). IEEE, 2013: 931-937.
[10]	李柏. 复杂约束下自动驾驶车辆运动规划的计算最优控制方法研究[D]. 杭州: 浙江大学, 2018.
	LI Bai. Research on computational optimal control methods for autonomous vehicle motion planning problems with complicated constraints[D]. Hangzhou: Zhejiang University, 2018. (in Chinese)
[11]	Bojarski M, Del Testa D, Dworakowski D, et al. End to end learning for self-driving cars[Z]. arXiv preprint arXiv: 1604. 07316, 2016.
[12]	Wang H, Kearney J, Atkinson K. Arc-length parameterized spline curves for real-time simulation[C]// Proc. 5th Int’l Conf Curves Surfaces. 2002: 387396.
[13]	TAKAHASHI Hongo. Local path planning and motion control for AGV in positioning[C]// Autonomous Mobile Robots Appl, IEEE, 2001: 392-397.
[14]	Polack P, Altche F, D'Andrea-Novel B, et al. The Kinematic bicycle model: A consistent model for planning feasible trajectories for autonomous vehicles?[C]// 2017 IEEE Intell Vehi Symp (IV). IEEE, 2017.
[15]	工业和信息化部, 公安部, 交通运输部. 关于印发《智能网联汽车道路测试管理规范(试行)》[J]. 信息技术与网络安全, 2018,37(4):99.
	Ministry of Industry and Information Technology, Ministry of Public Security, Ministry of Transportation. The issuance of "Management Specifications for Road Testing of Intelligent Connected Vehicles (Trial)"[J]. Info Tech Network Security, 2018,37(4):99. (in Chinese)