考虑驾驶风格与车辆交互关系的自由换道决策模型

龙雪琴; 毛健旭; 翟曼溶; 王远泽

doi:10.3963/j.jssn.1674-4861.2025.01.013

考虑驾驶风格与车辆交互关系的自由换道决策模型

doi: 10.3963/j.jssn.1674-4861.2025.01.013

长安大学运输工程学院西安 710064

基金项目:

陕西省自然科学基础研究计划 2024JC-YBMS-338

陕西省重点研发计划项目 2023-YBGY-138

详细信息

作者简介:
龙雪琴（1982—），博士，副教授. 研究方向：出行行为. E-mail：xqlong@chd.edu.cn

通讯作者:
龙雪琴（1982—），博士，副教授. 研究方向：出行行为. E-mail：xqlong@chd.edu.cn

中图分类号: U491
计量
- 文章访问数: 11
- HTML全文浏览量: 4
- PDF下载量: 0
- 被引次数: 0
出版历程
- 收稿日期: 2024-07-06
- 网络出版日期: 2025-06-27

A Free Lane-changing Decision Model Considering the Driving Style and the Interaction with Peripheral Vehicles

College of Transportation Engineering, Chang'an University, Xi'an 710064, China

摘要

摘要: 换道车辆与周围车辆之间的交互关系会对换道决策产生影响。为此基于换道效用构建了1种融合驾驶风格以及交互关系的换道决策模型。基于极端梯度提升树（extreme gradient boosting，XGBoost）算法和基于密度的聚类（density-based spatial clustering of applications with noise，DBSCAN）方法，将驾驶人的短时驾驶风格分为保守型、一般型、激进型这3类；根据换道车辆与目标车道后车之间的交通冲突，结合换道持续时间判断轨迹的时空重合点，划分车辆之间的交互关系。构建了速度提升、空间安全、时间安全3个维度的效用量化模型，基于最大信息系数（maximum information coefficients，MIC）计算各类驾驶人不同交互关系下3种子效用的权重，构建换道决策总效用模型。根据历史数据计算换道驾驶人与车道保持驾驶人的总效用，聚类得到不同驾驶风格、不同交互关系下的换道效用阈值，提出车辆换道决策的规则。模型准确率检验结果表明：无论是哪种类型的驾驶人，考虑交互关系的模型准确率明显高于不考虑交互关系的准确率，说明交互关系是影响换道决策的重要因素；分别采用径向基（radial basis function，RBF）神经网络与XGBoost方法进行换道行为预测，对于保守型、一般型、激进型驾驶人，RBF方法的准确率分别为0.885、0.820、0.813，XGBoost方法的准确率为0.954、0.902、0.900，可见人工智能模型虽然对各类驾驶人的换道决策行为预测精度都较高，但对人数占比较大的一般型与激进型驾驶人的预测准确率低于本文提出的决策模型（0.921，0.923）。此外，对换道效用进行了非参数检验，检验结果进一步证明了模型的合理性。
- 交通工程 /
- 换道决策 /
- 驾驶风格 /
- 交互关系 /
- 综合效用
Abstract: The interactions between lane-changing vehicle and peripheral vehicles will influence the lane-changing decision behavior. In response to this, a lane-changing decision model that integrates driving styles and interactions is developed based on the lane-changing utility. Utilizing the extreme gradient boosting (XGBoost) algorithm and the density-based spatial clustering of applications with noise (DBSCAN) clustering method, drivers' short-term driving styles are categorized into conservative, typical, and aggressive types. Based on traffic conflicts of trajectories, the spatiotemporal overlap points between vehicles are determined to classify interactions. A utility quantification model is constructed across three dimensions: speed enhancement, spatial safety, and temporal safety. The weights of three sub-utilities are calculated based on MIC, then a total utility model for lane-changing decision is established. Using historical data, the total utilities of lane-changing drivers are computed in comparison to that of lane-keeping drivers, resulting in the identification of lane-changing utility thresholds. Thereby rules for lane-changing decision are formulated. The result of model's accuracy indicates that the model considering interactions significantly outperforms the model that does not, underscoring the importance of interactions in lane-changing decision. Predictions of lane-changing behavior are conducted using the radial basis function (RBF) model and XGBoost model. For conservative, typical, and aggressive drivers, accuracies of the RBF method are 0.885, 0.820, and 0.813, while the XGBoost method achieves accuracies of 0.954, 0.902, and 0.900. AI models demonstrate high predictive accuracies for lane-changing decision across all driver types. However, the predictive accuracies for the typical and aggressive drivers are lower than that of the decision model proposed in this paper (0.921 and 0.923, respectively). Additionally, non-parametric statistical tests of lane-changing utilities further validate the rationality of the model.
- traffic engineering /
- lane change decision /
- driving style /
- interaction relationship /
- comprehensive

HTML全文

图 1 highD数据集坐标系

Figure 1. Coordinate system of the highD dataset

下载: 全尺寸图片幻灯片

图 2 换道时间分布图

Figure 2. Distribution of lane-change time

下载: 全尺寸图片幻灯片

图 3 目标车辆与周边车辆位置关系图

Figure 3. Position diagram between target vehicle and surrounding vehicle

下载: 全尺寸图片幻灯片

图 4 目标车辆与目标车道后车交互关系示意图

Figure 4. Diagrammatic sketch of Interaction relationship between target vehicle and following vehicle in the target lane

下载: 全尺寸图片幻灯片

图 5 换道决策过程流程图

Figure 5. Flow chart of the lane-change decision process

下载: 全尺寸图片幻灯片

图 6 换道效用统计图

Figure 6. Statistical gragh of lane-change utility

下载: 全尺寸图片幻灯片

图 7 换道效用克鲁斯卡尔-沃利斯检验

Figure 7. Kruskal-Wallis test of lane-change utility

下载: 全尺寸图片幻灯片

表 1 换道决策影响因素特征重要性排名

Table 1. Ranking of characteristic importance of influencing factors of lane-change decision

特征名称	特征含义	特征重要性	排名
T_TTC	与前车的碰撞时间	0.352	1
V_{x_TF}	目标车道后车x方向的速度	0.271	2
D_TP	与目标车道前车的纵向距离	0.096	3
V_{x_P}	前车x方向的速度	0.034	4
A_x	本车x方向的加速度	0.031	5
A_{x_TP}	目标车道前车x方向的加速度	0.029	6
V_{y_P}	前车y方向的速度	0.029	7
D_P	与前车的纵向距离	0.027	8
V_{x_TP}	目标车道前车x方向的速度	0.026	9
V_{y_TP}	目标车道前车y方向的速度	0.026	10
A_{y_TP}	目标车道前车y方向的加速度	0.026	11
A_{x_P}	前车x方向的加速度	0.018	12
A_{y_P}	前车y方向的加速度	0.018	13
V_x	本车x方向的速度	0.013	14
V_{y_TF}	目标车道后车y方向速度	0.007	15

下载: 导出CSV

表 2 聚类均值

Table 2. Cluster means

分类	T_TTC/s	D_TP/m	V_x /(m/s)	A_x /(m/s²)	V_y /(m/s)
-1	324.654	216.796	30.701	0.093	0.603
0	198.531	202.439	29.436	0.279	0.636
1	39.560	234.438	35.323	0.098	0.628

下载: 导出CSV

表 3 各类驾驶人不同交互环境下效用模型MIC值

Table 3. Values of MIC of benifit model in different interactive environments of various drivers

交互关系	保守型			一般型			激进型
交互关系	E_v	E_{s_e}	E_{s_t}	E_v	E_{s_e}	E_{s_t}	E_v	E_{s_e}	E_{s_t}
无交互	0.255	0.076	0.237	0.193	0.217	0.292	0.174	0.141	0.275
主动	0.306	0.110	0.487	0.221	0.514	0.083	0.267	0.284	0.007
被动	0.473	0.123	0.107	0.209	0.245	0.240	0.301	0.315	0.944

下载: 导出CSV

表 4 各类驾驶人不同交互环境下权重系数

Table 4. Weight coefficients in different interactive environments of various drivers

交互关系	保守型			一般型			激进型
交互关系	α^(i)(j)	β^(i)(j)	γ^(i)(j)	α^(i)(j)	β^(i)(j)	γ^(i)(j)	α^(i)(j)	β^(i)(j)	γ^(i)(j)
无交互	0.449	0.134	0.417	0.275	0.309	0.416	0.295	0.239	0.466
主动	0.339	0.122	0.539	0.270	0.628	0.101	0.478	0.509	0.013
被动	0.672	0.175	0.152	0.301	0.353	0.346	0.193	0.202	0.605

下载: 导出CSV

表 5 阈值计算结果

Table 5. Results of the threshold

类型	保守型	一般型	激进型
无交互	0.626	0.627	0.592
主动	-0.033	-0.21	0.086
被动	-0.139	-0.072	0.003

下载: 导出CSV

表 6 评价指标结果

Table 6. Results of the evaluation indicators

指标	考虑交互关系	不考虑交互关系
精确率	0.862	0.804
召回率	0.781	0.677
准确率	0.844	0.779
F1-score	0.819	0.735

下载: 导出CSV

表 7 各类驾驶人换道决策识别准确率

Table 7. Identification accuracy of various drivers

驾驶风格	无交互	主动	被动	加权平均值
驾驶风格	无交互	主动	被动	考虑交互	不考虑交互
保守型	0.804	0.764	0.728	0.791	0.736
一般型	0.944	0.795	0.773	0.921	0.765
激进型	0.942	0.770	0.833	0.923	0.801

下载: 导出CSV

表 8 RBF与XGBoost各类驾驶人识别准确率

Table 8. Recognition accuracy of RBF and XGBoost for different drivers

驾驶风格	RBF	XGBoost
保守型	0.885	0.954
一般型	0.820	0.902
激进型	0.813	0.900

下载: 导出CSV

表 9 换道驾驶人与不换道驾驶人换道效用的曼-惠特尼检验显著性

Table 9. Significances of the Mann-Whitney test of the lane-change benifit between drivers with and without lane change

换道效用	保守型			一般型			激进型
换道效用	无交互	主动	被动	无交互	主动	被动	无交互	主动	被动
E	0.014*	0.010*	0.057	0*	0.002*	0.020*	0*	0.002*	0.003*
E_v	0.758	0.019*	0.057	0*	0.004*	0.051	0.488	0.736	0.336
E_{s_e}	0.931	0.010*	0.229	0.021*	0.004*	0.109	0*	0.011*	0.633
E_{s_t}	0.068	0.610	0.857	0*	0.433	0.076	0*	0.507	0.010*
*表示在95%的置信区间下显著。

下载: 导出CSV

参考文献(26)

[1]	GIPPS P G. A model for the structure of lane-changing decisions[J]. Transportation Research Part B: Methodological, 1 986, 20(5): 403-414.
[2]	KESTING A, TREIBER M, HELBING D. General lane-changing model MOBIL for car-following models[J]. Transportation Research Record, 2007, 1999(1): 86-94. doi: 10.3141/1999-10
[3]	杨达, 吕蒙, 戴力源, 等. 车联网环境下自动驾驶车辆车道选择决策模型[J]. 中国公路学报, 2022, 35(4): 243-255. doi: 10.3969/j.issn.1001-7372.2022.04.020 YANG D, LYU M, DAI L Y, et al. Decision model for lane selection of autonomous vehicles in the vehicle environment[J]. Highway Journal of China, 2022, 35(4): 243-255. (in Chinese) doi: 10.3969/j.issn.1001-7372.2022.04.020
[4]	曲大义, 张可琨, 顾原, 等. 自动驾驶车辆换道决策行为分析及分子动力学建模[J]. 吉林大学学报(工学版), 2024, 54(3): 700-710. QU D Y, ZHANG K K, GU Y, et al. Analysis of lane change decision behavior and molecular dynamics modeling of autonomous vehicles[J]. Journal of Jilin University(Engineering Edition), 2024, 54(3): 700-710. (in Chinese)
[5]	邓建华, 冯焕焕. 基于换道决策机理的多车道元胞自动机模型[J]. 交通运输系统工程与信息, 2018, 18(3): 68-73. DENG J H, FENG H H. Multi-lane cellular automaton model based on lane change decision mechanism[J]. Transportation System Engineering and Information, 2018, 18(3): 68-73. (in Chinese)
[6]	秦雅琴, 王锦锐, 谢济铭, 等. 基于动态行车安全间距的交织区换道模型[J]. 安全与环境学报, 2023, 23(6): 1926-1934. QIN Y Q, WANG J R, XIE J M, et al. Interwoven area lane changing model based on dynamic traffic safety distance[J]. Journal of Safety and the Environment, 2023, 23(6): 1926-1934. (in Chinese)
[7]	TOLEDO T, KOUTSOPOULOS H N, BEN-AKIVA M E. Modeling integrated lane-changing behavior[J]. Journal of the Transportation Research Board, 2003(1857): 30-38
[8]	李林恒, 甘婧, 曲栩, 等. 智能网联环境下基于安全势场理论的车辆换道模型[J]. 中国公路学报, 2021, 34(6): 184-195. LI L H, GAN J, QU X, et al. Vehicle lane change model based on safety potential field theory in an intelligent connected environment[J]. Highway Journal of China, 2021, 34(6): 184-195. (in Chinese)
[9]	郭海兵, 曲大义, 洪家乐, 等. 基于效用理论的车辆换道交互行为及决策模型[J]. 科学技术与工程, 2020, 20(29): 12185-12190. doi: 10.3969/j.issn.1671-1815.2020.29.053 GUO H B, QU D Y, HONG J L, et al. Vehicle lane change interaction behavior and decision model based on utility theory[J]. Science, Technology and Engineering, 2020, 20(29): 12185-12190. (in Chinese) doi: 10.3969/j.issn.1671-1815.2020.29.053
[10]	陆春意, 何赏璐, 高彬彬, 等. 基于博弈论的自动驾驶驶离专用车道换道决策模型[J]. 交通信息与安全, 2024, 42(4): 144-153, 174. LU C Y, HE S L, GAO B B, et al. The decision model of autonomous driving leaving the dedicated lane based on game theory[J]. Traffic Information and Safety, 2024, 42(4): 144-153, 174. (in Chinese)
[11]	SCHUBERT R, WANIELIK G. A unified Bayesian approach for object and situation assessment[J]. Intelligent Transportation Systems Magazine, IEEE Intelligent Transportation Systems Magazine, 2011, 3(2): 6-19. doi: 10.1109/MITS.2011.941331
[12]	谷新平, 韩云鹏, 于俊甫. 基于决策机理与支持向量机的车辆换道决策模型[J]. 哈尔滨工业大学学报, 2020, 52(7): 111-121. GU X P, HAN Y P, YU J F. Vehicle lane change decision model based on decision mechanism and support vector machine[J]. Journal of Harbin Institute of Technology, 2020, 52(7): 111-121. (in Chinese)
[13]	徐兵, 刘潇, 汪子扬, 等. 采用梯度提升决策树的车辆换道融合决策模型[J]. 浙江大学学报(工学版), 2019, 53(6): 1171-1181. XU B, LIU X, WANG Z Y, et al. A vehicle lane change fusion decision model using a gradient lift decision tree[J]. Journal of Zhejiang University (Engineering Edition), 2019, 53(6): 1171-1181. (in Chinese)
[14]	张鑫辰, 张军, 刘元盛, 等. 改进深度Q网络的无人车换道决策算法研究[J]. 计算机工程与应用, 2022, 58(7): 266-275. ZHANG X C, ZHANG J, LIU Y S, et al. Research on improving the decision algorithm of unmanned vehicle in deep Q network[J]. Computer Engineering and Application, 2022, 58(7): 266-275. (in Chinese)
[15]	侯海晶, 金立生, 关志伟, 等. 驾驶风格对驾驶行为的影响[J]. 中国公路学报, 2018, 31(4): 18-27. HOU H J, JIN L S, GUAN Z W, et al. Effect of driving style on driving behavior[J]. Highway Journal of China, 2018, 31(4): 18-27. (in Chinese)
[16]	冯焕焕, 邓建华, 葛婷. 引入驾驶风格的熵权法多属性换道决策模型[J]. 交通运输系统工程与信息, 2020, 20(2): 139-144. FENG H H, DENG J H, GE T. The entropy-weight decision model of driving style is introduced[J]. Transportation System Engineering and Information, 2020, 20(2): 139-144. (in Chinese)
[17]	李珂, 韩同群, 杨正才, 等. 基于驾驶人心理风险场模型的个性化换道决策方法[J]. 交通信息与安全, 2023, 41(6): 32-41. doi: 10.3963/j.jssn.1674-4861.2023.06.004 LI K, HAN T Q, YANG Z C, et al. A personalized lane change decision method based on driver's psychological risk field model[J]. Journal of Transport Information and Safety, 2023, 41(6): 32-41. (in Chinese) doi: 10.3963/j.jssn.1674-4861.2023.06.004
[18]	曲大义, 陈文娇, 杨万三, 等. 车辆换道交互行为分析和建模[J]. 公路交通科技, 2016, 33(6): 88-94. QYU D Y, CHEN W J, YANG W S, et al. Analysis and modeling of vehicle lane change interaction behavior[J]. Highway Traffic Technology, 2016, 33(6): 88-94. (in Chinese)
[19]	杨志强, 朱家伟, 穆蕾, 等. 基于高斯混合隐马尔科夫模型的自由换道识别[J]. 计算机系统应用, 2022, 31(8): 388-394. YANG Z Q, ZHU J W, MU L, et al. Free-change identification based on Gaussian mixed hidden Markov model[J]. Application of Computer Systems, 2022, 31(8): 388-394. (in Chinese)
[20]	汪猛. 基于高精度数据的换车道行为特征研究[D]. 长沙: 湖南大学, 2017. WANG M. Study on the behavior characteristics of lane changing based on high-precision data[D]. Changsha: Hunan University, 2017. (in Chinese)
[21]	董俊一. 考虑驾驶风格的智能驾驶换道决策模型研究[D]. 长春: 吉林大学, 2022. DONG J Y. Study of intelligent driving considering driving style[D]. Changchun: Jilin University, 2022. (in Chinese)
[22]	ALI Y, ZHENG Z, HAQUE M M, et al. A game theory-based approach for modelling mandatory lane-changing behaviour in a connected environment[J]. Transportation Research Part C: Emerging Technologies, 2019, 106: 220-242.
[23]	孙一. 考虑自车与交通车行为交互的自动驾驶换道轨迹规划策略研究[D]. 长春: 吉林大学, 2023. SUN Y. Research on lane changing trajectory planning strategy of autonomous driving considering the behavioral interaction between self-vehicle and traffic vehicle[D]. Changchun: Jilin University, 2023. (in Chinese)
[24]	李高伟, 傅成红, 高良鹏. 人机混驾交通流跟驰特性建模与仿真研究[J/OL]. (2024-04-10)[2024-06-22]. http://kns.cnki.net/kcms/detail/42.1824.U.20240409.1159.030.html. LI G W, FU C H, GAO L P. Modeling and simulation of man-driving traffic flow and characteristics[J/OL]. (2024-04-10)[2024-06-22]. http://kns.cnki.net/kcms/detail/42.1824.U.20240409.1159.030.html.
[25]	洪家乐, 曲大义, 贾彦峰, 等. 基于驾驶人反应特性的车辆跟驰行为及模型[J]. 青岛理工大学学报, 2021, 42(4): 108-114, 142. HONG J L, QU D Y, JIA Y F, et al. Vehicle following behavior and model based on driver reaction characteristics[J]. Journal of Qingdao University of Technology, 2021, 42(4): 108-114, 142. (in Chinese)
[26]	朱煦晗. 智能汽车换道决策与轨迹规划算法研究[D]. 长春: 吉林大学, 2021. ZHU X H. Research on lane change decision and trajectory planning algorithm for intelligent vehicles[D]. Changchun: Jilin University, 2021. (in Chinese)