案例推理机制下智能汽车换道避障轨迹规划方法

doi:10.16265/j.cnki.issn1003-3033.2025.07.0122

摘要/Abstract

摘要：

为解决智能汽车换道避障轨迹规划难以兼顾性能优化与实时性的问题,提出基于案例推理(CBR)机制的智能汽车换道避障轨迹规划方法。利用五次多项式轨迹参数离线优化得到典型场景下的最优轨迹,形成状态量与决策量相对应的最优案例库,探明新案例场景下最优轨迹修正规律,建立轨迹关键参数更新函数,从而在相似案例场景下直接调用现有案例轨迹,在新案例场景下根据车速、位置等状态量在线更新换道避障轨迹,并进行案例库更新。对比测试结果表明:所提出方法与五次多项式在线优化方法的轨迹性能基本一致,且明显优于A^*和快速随机探索树星(RRT^*)方法。其中,所提出方法与五次多项式在线优化方法在相似案例场景下,所得轨迹最大曲率均为0.001 2 m^-1,最大侧向加速度均为1.245 7 m/s²;在新案例场景下,两者轨迹最大曲率均为0.001 4 m^-1,而最大侧向加速度分别为1.194 9和1.136 5 m/s²。同时,所提出方法与五次多项式在线优化、A^*和RRT^*这3种对比方法相比,具有最小的规划耗时,新案例场景下的规划耗时为0.001 1 s,而相似案例场景下的规划耗时仅为0.000 1 s。

关键词: 案例推理(CBR), 智能汽车, 换道避障, 轨迹规划, 区别度

Abstract:

To solve the problem of difficulty in balancing trajectory optimization and real-time performance in lane change and obstacle avoidance trajectory planning for intelligent vehicles, a CBR mechanism based intelligent vehicle lane changing and obstacle avoidance trajectory planning method was proposed. The optimal trajectories in typical scenarios were obtained by offline optimization of the fifth-degree polynomial trajectory parameters, forming an optimal case library consisting of the state variables and decision variables. The optimal trajectory correction rule in the new case scenario was explored, and a trajectory key parameter update function was established. Thus, existing case trajectories were directly called in similar case scenarios, and the lane changing and obstacle avoidance trajectories were updated according to state variables such as vehicle speed and position in new case scenarios, thus the case library was updated. The comparative test results show that the trajectory performance of proposed method is basically consistent with that of the fifth-degree polynomial online optimization method, and is significantly better than that of A^* and Rapidly-exploring Random Trees-star(RRT^*) methods. Specifically, the maximum curvature of the trajectories from the proposed method and the fifth-degree polynomial online optimization method are both 0.001 2 m^-1, and the maximum lateral acceleration of the trajectories from the two methods are both 1.245 7 m/s² in the similar case scenario. While in the new case scenario, the maximum curvature of the trajectories from the two methods are both 0.001 4 m^-1, and the maximum lateral acceleration is 1.194 9 and 1.1365 m/s², respectively. Meanwhile, compared with the three comparative methods of the fifth-degree polynomial online optimization, A^* and RRT^*, the proposed method has the smallest planning time. The planning time in the new case scenario and the similar case is only 0.001 1 and 0.000 1 s, respectively.

Key words: case-based reasoning(CBR), intelligent vehicles, lane change and obstacle avoidance, trajectory planning, difference degree

中图分类号:

X951交通运输安全

房熙博, 宁一高, 赵轩, 周猛. 案例推理机制下智能汽车换道避障轨迹规划方法[J]. 中国安全科学学报, 2025, 35(7): 209-217.

FANG Xibo, NING Yigao, ZHAO Xuan, ZHOU Meng. Lane change and obstacle avoidance trajectory planning method for intelligent vehicles using case-based reasoning mechanism[J]. China Safety Science Journal, 2025, 35(7): 209-217.

图/表 12

图1

图2

图3

图4

图5

表1

图6

图7

图8

表2

图9

表3

参考文献 16

[1]	RONG Donglei, JIN Sheng, YANG Chengcheng. Connected and autonomous vehicle trajectory planning considering communication delay[J]. IEEE Transactions on Vehicular Technology, 2024, 73(9): 12 668-12 681.
[2]	姚佼, 杨承逸, 杨媛媛, 等. 基于安全势场理论的高速公路交通事故下协同换道引导策略[J]. 中国安全科学学报, 2024, 34(7): 71-82.
	YAO Jiao, YANG Chengyi, YANG Yuanyuan, et al. Cooperative lane change guidance strategy in freeway traffic accident area based on safety potential field theory[J]. China Safety Science Journal, 2024, 34(7): 71-82.
[3]	ALI Y. Autonomous vehicle lane-changing dynamics and impact on the immediate follower[J]. Analytic Methods in Accident Research, 2025, 46(6): DOI: 10.1016/j.amar.2025.100388.
[4]	XIN Long, KONG Yiting, LI S E, et al. Enable faster and smoother spatio-temporal trajectory planning for autonomous vehicles in constrained dynamic environment[J]. Proceedings of the Institution of Mechanical Engineers, Part D: Journal of Automobile Engineering, 2021, 235(4): 1101-1112.
[5]	XU Biao, WANG Guan, YANG Zeyu, et al. Multi-vehicle collaborative trajectory planning in unstructured conflict areas based on v-hybrid A^*[J]. IEEE Transactions on Intelligent Transportation Systems, 2024, 25(9): 12 722-12 735.
[6]	LI Bai, KONG Qi, ZHANG Youmin, et al. On-road trajectory planning with spatio-temporal RRT^* and always-feasible quadratic program[C]. Proceeding in 16^th International Conference on Automation Science and Engineering, 2020: 942-947.
[7]	YAN Yongjun, WANG Jinxiang, WANG Yan, et al. A cooperative trajectory planning systems based on the passengers' individual preferences of aggressiveness[J]. IEEE Transactions on Vehicular Technology, 2025, 72(1): 395-406.
[8]	LIU Yonggang, ZHOU Bobo, WANG Xiao, et al. Dynamic lane-changing trajectory planning for autonomous vehicles based on discrete global trajectory[J]. IEEE Transactions on Intelligent Transportation Systems, 2024, 23(7): 8513-8527.
[9]	谭睿璞, 张文德, 陈圣群, 等. 异质信息环境下基于案例推理的应急决策方法[J]. 控制与决策, 2020, 35(8): 1966-1976.
	TAN Ruipu, ZHANG Wende, CHEN Shengqun, et al. Emergency decision-making method based on case-based reasoning in heterogeneous information environment[J]. Control and Decision, 2020, 35(8): 1966-1976.
[10]	LI Yang, LI Linbo, NI Daiheng, et al. Automatic lane-changing trajectory planning: from self-optimum to local-optimum[J]. IEEE Transactions on Intelligent Transportation Systems, 2022, 23(11): 21 004-21 014.
[11]	牛国臣, 李文帅, 魏洪旭. 基于双五次多项式的智能汽车换道轨迹规划[J]. 汽车工程, 2021, 43(7): 978-986, 1 004.
	NIU Guochen, LI Wenshuai, WEI Hongxu. Intelligent vehicle lane changing trajectory planning based on double quantic polynomials[J]. Automotive Engineering, 2021, 43(7): 978-986, 1 004.
[12]	YUE Ming, YANG Lu, ZHANG Hongzhi, et al. Automated hazard escaping trajectory planning/tracking control framework for vehicles subject to tire blowout on expressway[J]. Nonlinear Dynamics, 2019, 98: 61-74.
[13]	ZHANG Yuelou, CHEN Lingshan, LI Ning. Improved quantic polynomial autonomous vehicle lane-change trajectory planning based on hybrid algorithm optimization[J]. World Electric Vehicle Journal, 2025, 16(5): DOI: 10.3390/wevj16050244.
[14]	洪亮, 陈志豪, 李梁, 等. 混杂工况下面向VRU安全的主动转向控制策略[J]. 中国安全科学学报, 2024, 34(12): 84-93.
	HONG Liang, CHEN Zhihao, LI Liang, et al. Active steering control strategy for VRU safety under mixed conditions[J]. China Safety Science Journal, 2024, 34(12): 84-93.
[15]	KABIR R, WATANOBE Y, ISLAM M R, et al. Enhanced robot of A-star algorithm for robotic path planning[J]. Sensors, 2024, 24(5): DOI: 10.3390/s24051422.
[16]	ZHANG Shi, CUI Rongxin, YAN Weisheng, et al. RA-RRTV^: risk-averse RRT^ with local vine expansion for path planning in narrow passages under localization uncertainty[J]. IEEE Robotics and Automation Letters, 2025, 10(2): 2072-2079.

状态量			决策量		状态量			决策量		状态量			决策量
${{v}_{x,}}_{m}$/ (m·s^-1)	${{v}_{x}}_{,o}$/ (m·s^-1)	y_r/ m	S/ m	T	${{v}_{x,}}_{m}$/ (m·s^-1)	${{v}_{x}}_{,o}$/ (m·s^-1)	y_r/ m	S/ m	T	${{v}_{x,}}_{m}$ (m·s^-1)	${{v}_{x}}_{,o}$/ (m·s^-1)	y_r/ m	S/ m	T
40	20	-0.625	16.44	1	80	60	-0.625	9.02	16	120	20	-0.625	68.12	31
40	20	0	21.58	2	80	60	0	18.02	17	120	20	0	82.06	32
40	20	0.625	41.26	3	80	60	0.625	25.12	18	120	20	0.625	91.66	33
60	20	-0.625	24.65	4	100	20	-0.625	51.61	19	120	40	-0.625	59.41	34
60	20	0	38.63	5	100	20	0	68.29	20	120	40	0	65.66	35
60	20	0.625	46.44	6	100	20	0.625	78.36	21	120	40	0.625	71.33	36
60	40	-0.625	9.21	7	100	40	-0.625	39.81	22	120	60	-0.625	37.11	37
60	40	0	13.45	8	100	40	0	51.23	23	120	60	0	49.26	38
60	40	0.625	19.36	9	100	40	0.625	64.44	24	120	60	0.625	56.72	39
80	20	-0.625	39.11	10	100	60	-0.625	23.62	25	120	80	-0.625	21.01	40
80	20	0	53.93	11	100	60	0	34.17	26	120	80	0	32.86	41
80	20	0.625	61.15	12	100	60	0.625	40.04	27	120	80	0.625	42.94	42
80	40	-0.625	21.16	13	100	80	-0.625	8.91	28	120	100	-0.625	9.78	43
80	40	0	35.98	14	100	80	0	17.12	29	120	100	0	16.45	44
80	40	0.625	45.16	15	100	80	0.625	26.31	30	120	100	0.625	16.03	45

评价指标	文中方法	五次多项式	A^*算法	RRT^*算法
轨迹最大曲率/m^-1	0.001 2	0.001 2	0.165 5	0.098 0
与障碍车的最小距离/m	0.305 3	0.305 3	0.193 1	0.764 9
最大侧向加速度/(m·s^-2)	1.245 7	1.245 7	3.114 5	2.711 4
规划耗时/s	0.000 1	0.001 3	0.288 3	0.087 0

评价指标	文中方法	五次多项式	A^*算法	RRT^*算法
轨迹最大曲率/m^-1	0.001 4	0.001 4	0.165 5	0.085 0
与障碍车的最小距离/m	0.163 0	0.141 4	0.246 6	0.291 9
最大侧向加速度/(m·s^-2)	1.194 9	1.136 5	3.166 5	2.044 7
规划耗时/s	0.001 1	0.002 8	0.289 8	0.056 0