A Comprehensive Evaluation Model of Lane-based Risk for Urban Intersections Based on Bayesian Inference and XGBoost
-
摘要: 针对城市信号控制交叉口车道风险研究不足及复杂交互带来的不确定性等问题，建立了基于贝叶斯推理与XGBoost的车道风险综合评价模型。基于西安市吉祥村、明光路和青松路这3个交叉口的交通视频数据，从时间逼近和空间逼近这2个维度构建了2个新的风险评价集，选取后侵入时间、最大速度、距离差和速度差作为核心指标，用以捕捉交叉口内的动态风险特征；并引入贝叶斯推理构建概率性评价方法，以解决复杂交互中的不确定性问题。随后进行XGBoost模型的SHAP值理论和Logistic回归，探究影响车道风险等级的特征重要程度和显著性。结果表明：①建立的车道风险综合评价模型在评估非机动车-汽车、行人-汽车与行人-非机动车等3类交互冲突时，识别中等和高风险的性能优于基准模型，特别是在极度危险交互的判定上更为合理，避免了基准模型的高估问题。②在常见的非机动车-汽车交互、行人-汽车交互和行人-非机动车交互中，仅有少部分为极度危险情况，但仍存在较多的中等风险，总占比分别为29.7%，20.8%，34.3%。③交叉口不同车道的风险存在显著差异，第1车道相较于第2、3、4车道更容易发生交通冲突。④在3类交互中，车道风险主要受速度、加速度和流量影响。非机动车-汽车交互在第1车道和隔离带宽度较小的路段风险最高，尤其是早高峰和右转车道。行人-汽车交互的车道风险因素集中于速度和流量，且第1车道风险较大。行人-非机动车交互中，较窄的非机动车道增加了冲突风险。Abstract: This study addresses the challenges of inadequate researches on lane-scale risk evaluation and the uncertainties in complex interactions at urban signal-controlled intersections. To this end, a comprehensive risk evaluation model is developed based on Bayesian inference and XGBoost. Specifically, this study is based on traffic video data from three intersections in Xi'an: Jixiang Village, Mingguang Road, and Qingsong Road. Two innovative risk-evaluation sets are constructed from the dimensions of temporal and spatial proximities, in which key indicators, including post-entrainment time, maximum speed, distance difference and speed difference are selected to capture dynamic risk characteristics of intersections. Further, Bayesian inference is used to develop a probabilistic evaluation method to address uncertainties in complex interactions at intersections. Next, SHAP value theory of the XGBoost model and Logistic regression are applied to analyze the significance and importance of factors influencing lane risk levels. The results show that: ①The proposed model outperforms baseline models in identifying medium and high-risk interactions. It also more accurately assesses extreme danger interactions, avoiding the overestimation observed in baseline models. ②Among the typical interactions, that between motor vehicle-bicycle, pedestrian-motor vehicle, and pedestrian-bicycle, only a small portion are classified as extreme risk, though medium-risk interactions account for 29.7%, 20.8%, and 34.3%, respectively. ③There are significant differences regarding the risk level across different lanes, with the first lane being more prone to traffic conflicts compared to the second, third, and fourth lanes. ④For all three interaction types, lane risk is mainly influenced by speed, acceleration, and traffic volume. In motor vehicle-bicycle interactions, the highest risk occurs in the first lane and on roads with narrow buffer zones, particularly during morning rush hours and on right-turn lanes. Pedestrian-motor vehicle interactions are primarily influenced by speed and traffic volume, with higher risks in the first lane. For pedestrian-bicycle interactions, narrower bicycle lanes increase the risk of conflicts.
-
Key words:
- traffic safety /
- lane risk /
- safety analysis /
- Bayesian inference /
- XGBoost
-
表 1 3类交互的统计信息
Table 1. Statistical information for three types of interactions
变量 类别 非机动车-汽车 行人-汽车 行人-非机动车 频数 占比//% 频数 占比//% 频数 占比//% 车道 第1车道 1 173 46.68 1 716 67.24 9 056 82.62 第2车道 363 14.44 281 11.01 748 6.82 第3车道 344 13.69 307 12.03 546 4.98 第4车道 633 25.19 248 9.72 611 5.57 右转 1 173 46.68 1 716 67.24 9 056 82.62 车道类型 直行 707 28.13 588 23.04 1 294 11.81 左转 633 25.19 248 9.72 611 5.57 车道宽度/m 3 2 344 98.27 1 365 53.49 3 764 34.34 3.5 169 6.73 1 187 46.51 7 197 65.66 隔离带宽度/m 0 714 28.41 693 27.16 1 816 16.57 2 1 799 71.59 1 859 72.84 9 145 83.43 2 714 28.41 693 27.16 1 816 16.57 非机动车道宽度/m 3 940 37.41 150 5.88 271 2.47 3.5 859 34.18 1 709 66.97 8 874 80.96 早高峰（07：30—09：30） 1 424 56.67 1 139 44.63 5 900 53.83 时间 午平峰（12：30—14：00） 612 24.35 783 30.68 2 744 25.03 晚高峰（17：00—18：30） 477 18.98 630 24.69 2 317 21.14 吉祥村交叉口 859 34.18 1 709 66.79 8 874 80.96 地点 明光路交叉口 940 37.41 150 5.88 271 2.47 青松路交叉口 714 28.41 693 27.16 1 816 16.57 
下载: 导出CSV
表 2 3种评价方法对比
Table 2. Comparison of three evaluation methods
交互类型 本文模型 基准模型1 基准模型2 非机动车-汽车 行人-汽车 行人-非机动车 非机动车-汽车 行人-汽车 行人-非机动车 非机动车-汽车 行人-汽车 行人-非机动车 安全 1 761（70.1%） 2 013（78.9%） 7 043（64.3%） 1 903（75.8%） 2 135（83.7%） 6 389（58.3%） 1 295（53.0%） 1 346（57.2%） 6 338（57.8%） 危险 746（29.7%） 530（20.8%） 3 756（34.3%） 475（18.9%） 324（12.7%） 3 540（32.3%） 984（40.3%） 632（26.9%） 3 686（33.6%） 极度危险 6（0.2%） 9（0.4%） 162（1.5%） 133（5.3%） 93（3.6%） 1 032（3.6%） 164（6.7%） 374（15.9%） 937（8.5%）
下载: 导出CSV
表 3 4种风险预测模型对比
Table 3. Comparison of four risk prediction models
模型 准确率/% 精确率/% 召回率/% F-1分数/% AUC/% XGBoost 91.51 94.17 62.58 75.19 93 GBDT 81.17 67.35 20.75 31.73 70 决策树 76.39 43.79 42.14 42.95 65 随机森林 81.83 65.88 34.14 44.98 79
下载: 导出CSV
表 4 3种交互类型Logistic回归表
Table 4. Three types of interaction Logistic regression table
影响因素 非机动车（1）-汽车（2） 行人（1）-汽车（2） 行人（1）-非机动车（2） B P B P B P 速度（2） 0.11 0.581 0.094 <0.01*** -0.075 <0.01*** 加速度（2） -0.02 0.739 -0.009 0.594 0.006 0.015 速度（1） 0.106 <0.01*** -0.077 0.524 -0.026 0.411 加速度（1） -0.09 0.123 0.01 0.517 0.001 0.721 流量（2） -0.02 <0.01*** 0.001 0.078* 0 0.739 流量（1） 0 0.921 0 0.836 0 0.016** 早晨（07：30—9：30） 0.379 0.034** 0.086 0.65 -0.065 0.353 中午（12：30—14：00） 1.145 <0.01*** 0.276 0.177 -0.154 0.051* 第1车道 0.836 <0.01*** -0.533 0.46 -0.031 0.849 第2车道 0.923 <0.01*** -0.283 0.274 -0.034 0.77 第3车道 0.915 <0.01*** -0.53 0.829 0.11 0.373 车道宽度3 m 0.48 0.066** -0.539 0.043** 0.021 0.892 隔离带宽度0 m -2.07 <0.01*** 0.665 0.024** -0.022 0.891 非机动车道宽度3 m -0.595 0.046** 0.685 0.189 -0.413 0.067* 注：“***”为99%水平下显著；“**”为95%水平下显著；“*”为90%水平下显著。
下载: 导出CSV
-
[1] CHEN W Q, WANG T, WANG Y J, et al. Lane-based distance-velocity model for evaluating pedestrian-vehicle interaction at non-signalized locations[J]. Accident Analysis & Prevention, 2022, 176: 106810. [2] MA Z J, MEI G, CUOMO S. An analytic framework using deep learning for prediction of traffic accident injury severity based on contributing factors[J]. Accident Analysis & Prevention, 2021, 160: 106322. [3] 王春勤, 牟瑞芳, 阮黄山. 基于灰云模型的城市道路交叉口交通冲突严重性研究[J]. 交通运输工程与信息学报, 2019, 17(3): 162-169.VUONG X C, MOU R F, NGUYEN H S. Study on traffic conflict severity of urban intersections based on grey cloud model[J]. Journal of Transportation Engineering and Information, 2019, 17(3): 162-169. (in Chinese) [4] WANG T, GE Y E, WANG Y J, et al. A novel model for real-time risk evaluation of vehicle-pedestrian interactions at intersections[J]. Accident Analysis & Prevention, 2024, 206: 107727. [5] 李杰, 曾叙砜, 孙领, 等. 全国道路安全水平的时空布局演化研究[J]. 中国安全科学学报, 2021, 31(12): 136-143.LI J, ZENG X F, SUN L, et al. Research on temporal and spatial evolution of road safety in China[J]. China Safety Science Journal, 2021, 31(12): 136-143. (in Chinese) [6] MOKHTARIMOUSAVI S, ANDERSON J C, AZIZINAMINI A, et al. Factors affecting injury severity in vehicle-pedestrian crashes: a day-of-week analysis using random parameter ordered response models and artificial neural networks[J]. International Journal of Transportation Science and Technology, 2020, 9(2): 100-115. doi: 10.1016/j.ijtst.2020.01.001 [7] 石琦, 方勇, 胡华, 等. 平面信号控制交叉口有轨电车与机动车交通冲突风险评价模型[J]. 中国安全生产科学技术, 2020, 16(增刊1): 147-151.SHI Q, FANG Y, HU H, et al. Risk assessment model of traffic conflict between tram and vehicle at signalized intersections[J]. Journal of Safety Science and Technology, 2020, 16(S1): 147-151. (in Chinese) [8] ALMODFER R, XIONG S W, FANG Z X, et. al. Quantitative analysis of lane-based pedestrian-vehicle conflict at a non-signalized marked crosswalk[J]. Transportation Research Part F: Traffic Psychology and Behaviour, 2016, 42(3): 468-478. [9] 郭延永, 刘攀, 吴瑶, 等. 基于冲突模型的非常规信号交叉口安全评价[J]. 中国公路学报, 2022, 35(1): 85-92. doi: 10.3969/j.issn.1001-7372.2022.01.008GUO Y Y, LIU P, WU Y, et al. Safety evaluation of unconventional signalized intersection based on traffic conflict extreme model[J]. China Journal of Highway and Transport, 2022, 35(1): 85-92. (in Chinese) doi: 10.3969/j.issn.1001-7372.2022.01.008 [10] JIANG C Z, QIU R, FU T, et al. Impact of right-turn channelization on pedestrian safety at signalized intersections[J]. Accident Analysis & Prevention, 2020, 136: 105399. [11] 朱顺应, 蒋若曦, 王红, 等. 机动车交通冲突技术研究综述[J]. 中国公路学报, 2020, 33(2): 15-33.ZHU S Y, JIANG R X, WANG H, et al. Review of research on traffic conflict techniques[J]. China Journal of Highway and Transport, 2020, 33(2): 15-33. (in Chinese) [12] GOVINDA L, RAJU M S K, SHANKAR K R. Pedestrian-vehicle interaction severity level assessment at uncontrolled intersections using machine learning algorithms[J]. Safety Science, 2022, 153: 105806. doi: 10.1016/j.ssci.2022.105806 [13] OLSZEWSKI P, DABKOWSKI P, SZAGALA P, et al. Surrogate safety indicator for unsignalised pedestrian crossings[J]. Transportation Research Part F: Traffic Psychology and Behaviour, 2020, 70: 25-36. doi: 10.1016/j.trf.2020.02.011 [14] 葛慧敏, 周礼军, 薄云钰, 等. 基于交通冲突理论的道路安全评价技术研究综述[J]. 交通信息与安全, 2022, 40(6): 12-21. doi: 10.3963/j.jssn.1674-4861.2022.06.002GE H M, ZHOU L J, BO Y Y, et al. A review of road safety evaluation techniques based on traffic conflict theories[J]. Journal of Transport Information and Safety, 2022, 40(6): 12-21. (in Chinese) doi: 10.3963/j.jssn.1674-4861.2022.06.002 [15] 吕能超, 王玉刚, 周颖, 等. 道路交通安全分析与评价方法综述[J]. 中国公路学报, 2023, 36(4): 183-201. doi: 10.3969/j.issn.1001-7372.2023.04.016LYU N C, WANG Y G, ZHOU Y, et al. Review of road traffic safety analysis and evaluation methods[J]. China Journal of Highway and Transport, 2023, 36(4): 183-201. (in Chinese) doi: 10.3969/j.issn.1001-7372.2023.04.016 [16] 褚昭明, 陈瑞祥, 刘金广. 城市道路无信号控制路段行人过街风险分级预警模型[J]. 交通信息与安全, 2023, 41(1): 53-61. doi: 10.3963/j.jssn.1674-4861.2023.01.006CHU Z M, CHEN R X, LIU J G. A model of risk classification and forewarning for pedestrian crossing behavior at unsignalized urban roadways[J]. Journal of Transport Information and Safety, 2023, 41(1): 53-61. (in Chinese) doi: 10.3963/j.jssn.1674-4861.2023.01.006 [17] 陆毅忱, 邹亚杰, 程凯, 等. 基于风险域的城市道路交叉口交通冲突分析方法[J]. 同济大学学报(自然科学版), 2021, 49(7): 941-948.LU Y C, ZOU Y J, CHENG K, et al. Analytical method of traffic conflict at urban road intersections based on risk region[J]. Journal of Tongji University (Natural Science), 2021, 49(7): 941-948. (in Chinese) [18] HU J C, LI X, CEN Y Q, et al. A roadside decision-making methodology based on deep reinforcement learning to simultaneously improve the safety and efficiency of merging zone[J]. IEEE Transactions on Intelligent Transportation Systems, 2022, 23(10): 18620-18631. doi: 10.1109/TITS.2022.3157910 [19] 胡立伟, 薛宇, 赵雪亭, 等. 驾驶人不同驾龄条件下公路交通风险致因耦合影响研究[J]. 安全与环境学报, 2023, 30(2): 1-9.HU L W, XUE Y, ZHAO X T, et al. Study on the coupling effect of road traffic risk causes under different driving ages[J]. Journal of Safety and Environment, 2023, 30(2): 1-9. (in Chinese) [20] XU C, ZHAO W., WANG C. An integrated threat assessment algorithm for decision-making of autonomous driving vehicles[J]. IEEE Transactions on Intelligent Transportation Systems, 2020, 21(6): 1-12. doi: 10.1109/TITS.2020.2995537 [21] NOH S. Decision-making framework for autonomous driving at road intersections: safeguarding against collision, overly conservative behavior, and violation vehicles[J]. IEEE Transactions on Industrial Electronics, 2018, 66(4): 3275-3286. [22] 郭延永, 刘攀, 吴瑶, 等. 基于贝叶斯多元泊松-对数正态分布的交通冲突模型[J]. 中国公路学报, 2018, 31(1): 101-109.GUO Y Y, LIU P, WU Y, et al. Traffic conflict model based on bayesian multivariate poisson-lognormal normal distribution[J]. China Journal of Highway and Transport, 2018, 31 (1): 101-109. (in Chinese) [23] ZHANG C B, ZHOU B, CHEN G J, et al. Quantitative analysis of pedestrian safety at uncontrolled multi-lane mid-block crosswalks in China[J]. Accident Analysis & Prevention, 2017, 108: 19-26. [24] AMINI R E, YANG K, ANTONIOU C. Development of a conflict risk evaluation model to assess pedestrian safety in interaction with vehicles[J]. Accident Analysis & Prevention, 2022, 175: 106773. -
点击查看大图
图(4) / 表(4)
计量
- 文章访问数: 91
- HTML全文浏览量: 41
- PDF下载量: 22
- 被引次数: 0