Analysis of SHRP2 Speeding Data

Project Description Phase 1: Speeding-related crashes continue to be a serious problem, and the proportion of speeding-related fatal crashes has changed little over a decade. Speeding is a complicated behavior that varies by driver and situation, and it also seems to be a common behavior, with most drivers reporting they drive over the speed limit at least some of the time. This dataset supported an initial exploration of the SHRP 2 data with regard to speeding. The objectives of that project were as follows: Identify, catalogue, and categorize SHRP2 data variables related to target research questions. Conduct test analyses using multiple approaches on a few cases and provide a detailed plan for an analysis of the complete sample to answer the target questions. Collect the SHRP2 data for these variables, code the variables, and produce a cleaned data file with corresponding data documentation. Analyze the data with regard to the speeding-related questions targeted for this study. Produce two final reports: one report on the findings from analysis of the speeding questions targeted, and a second report on the methods / process used to create the data file. Phase 2: Speeding continues to be a major factor in traffic crashes and fatalities. In 2018, 25% of all fatal crashes were identified as speeding-related. However, despite the importance of understanding it, studying speeding behavior poses many challenges. The Strategic Highway Research Program 2 (SHRP2) Naturalistic Driving Study has provided a unique dataset that has the potential to overcome some of these challenges and greatly enhance our understanding of speeding behavior. Battelle recently completed the first ever study using the SHRP2 speeding data (Contract No.: DTNH22-11-D-00229 / 0002). For the current project, the goal is to further develop the SHRP2 Speeding Database created during Phase 1 and conduct two new analyses to better understand drivers' speeding behavior. The primary addition of this new project to the existing SHRP2 Speeding Database will be the use of video data from drivers' vehicles, which will provide better data to define, identify, validate and predict speeding and aggressive driving behavior. In this follow-up study using the SHRP 2 Speeding data, Battelle will address the following target research questions on speeding and aggressive driving behavior: What is the role of certain pre-crash driver and situational factors in predicting speeding-related crashes (and near crashes)? What is the nature of speeding and aggressive driving behaviors in ambient traffic? Data Request Scope Phase 1: Video reductions for critical event and baseline epochs Size: All available epochs as of V1 of InSight Driver IDs, Driver demographics Size: All available records as of V1 of InSight Trip summaries Size: All available records in V1 of InSight with the following characteristics Trip Duration >= 15 minutes % CTRE Van Coverage >= 50% Max Speed >=25 mph Time Moving >= 50% * Trip Duration Trip time series sample Size: Trips selected by Battelle The primary study dataset consisted of the following number of trips selected by Battelle: Pennsylvania 16,344 trips Washington 22,974 trips Florida 67,163 trips Indiana 21,300 trips North Carolina 49,313 trips New York 69,543 trips Phase 2: Data for three types of driving epochs: 750 speeding related safety critical events identified by Battelle 1,532 base epochs, selected by Battelle from the existing Balanced Baselines and Additional Baselines 260 forward video trip segments identified by Battelle. Event details were provided for the safety critical events and baselines. The time series data and Navteq map data were provided for 60 seconds prior to the event for these groups. The time series data and Navteq roadway data, along with forward video, were provided for the duration of the forward video trip segments. In order to protect participant personally identifying information (PII), origin and destination buffers were employed. Position information and forward video, as applicable, were not provided for epochs that overlapped with these buffers. Position information cannot be released for crash events and forward video for crash events. Data Specification The variables included in this dataset can be found in the accompanying attached data dictionaries: Video reductions for critical event and baseline epochs Please see Event_Table_Data_Dictionary.pdf Driver IDs, Driver demographics Please see DriverDemographic_DataDictionary.pdf Trip summaries Please see Trip SummaryDataDictionary.pdf Trip time series sample Please see TimeSeriesData Dictionary.pdf Time series data was rectangularized at 1 Hz. Files where latitude or longitude were not available were excluded. Additional Analyses and Speeding Database Development This dataset consists of data reductions developed for the study of speeding in the SHRP2 Analysis of Speeding Data study, with new data developed in the current SHRP2 Speeding Data: Additional Analyses and Speeding Database Development study. The Analysis of SHPR2 Speeding Data study examined the problem of speeding using a sample of SHPR2 data from each of the six data collection sites included in the SHRP2 Naturalistic Driving Study. Time series data from each site were processed to identify Free-Flow Episodes (FFEs), in which the driver had opportunity to speed, and Speeding Episodes (SEs). Three data reductions that characterize driving behavior within the respective Trip, FFE, and SE were developed: (1) Trip-level reductions, (2) Free-Flow Episode (FFE) reductions, and (3) Speeding Episode (SE) reductions. Details describing how these datasets were produced and analyzed are provided in the Findings Report (Richard et al., 2018) and Methods Report (Brown & Richard, 2018) from that study. Two new data reductions were developed in the current study to address one of the study’s research questions: What is the nature of speeding and aggressive driving behaviors in ambient traffic? These reductions provide data that researchers can use to examine aggressive lane changes and tailgating while speeding. The data reduction and additional information can be found in the SHRP2 Speeding Data: Additional Analyses and Speeding Database Development zipped folder in the Files tab below.

项目描述
第一阶段：与超速相关的撞车事故至今仍是严峻的交通安全问题，十余年来，涉及超速的致命撞车事故占比几乎未有变化。超速行为是一种复杂的个体行为，因驾驶员个体与场景不同而存在差异，同时也是一种普遍现象——多数驾驶员自述曾至少在部分时段超速行驶。本数据集用于针对超速问题开展的战略公路研究计划2（Strategic Highway Research Program 2，简称SHRP2）数据初步探索研究。本项目目标如下：
1. 识别、整理并归类与目标研究问题相关的SHRP2数据变量；
2. 针对少量案例采用多种方法开展测试分析，并为完整样本的分析制定详细方案，以解答目标研究问题；
3. 收集对应变量的SHRP2数据，完成变量编码，并生成清洗后的数据文件及配套数据文档；
4. 针对本研究设定的超速相关问题开展数据分析；
5. 产出两份最终报告：一份为针对目标超速问题的分析结果报告，另一份为用于生成数据文件的方法与流程报告。

第二阶段：超速依旧是道路交通撞车事故与致死事故的核心影响因素。2018年，25%的致命撞车事故被认定为与超速相关。尽管理解超速行为至关重要，但对其开展研究仍面临诸多挑战。战略公路研究计划2（SHRP2）自然驾驶研究提供了独特的数据集，有望克服部分研究瓶颈，大幅加深我们对超速行为的认知。巴特尔（Battelle）近期已完成首项基于SHRP2超速数据的研究（合同编号：DTNH22-11-D-00229 / 0002）。本项目旨在进一步完善第一阶段构建的SHRP2超速数据库，并开展两项全新分析，以更深入地理解驾驶员的超速行为。本新增项目对现有SHRP2超速数据库的核心补充在于引入驾驶员车辆的视频数据，这将为定义、识别、验证及预测超速与激进驾驶行为提供更优质的数据支撑。在本次基于SHRP2超速数据的后续研究中，巴特尔将针对超速与激进驾驶行为的以下目标研究问题展开探索：
1. 撞车前的特定驾驶员个体与场景因素，在预测与超速相关的撞车事故（及险情）中发挥何种作用？
2. 常规交通场景下的超速与激进驾驶行为具有何种特征？

数据请求范围
第一阶段：
- 关键事件与基准时段的视频缩减数据：规模为InSight驾驶员ID V1版本发布时的所有可用时段
- 驾驶员ID与驾驶员人口统计学数据：规模为InSight行程摘要V1版本发布时的所有可用记录
- 行程摘要数据：规模为InSight V1版本中满足以下特征的所有可用记录：
- 行程时长≥15分钟
- CTRE货车覆盖占比≥50%
- 最高车速≥25英里每小时（mph）
- 行驶时间≥50%×行程时长
- 行程时序采样数据：规模为巴特尔选定的行程
本核心研究数据集包含巴特尔选定的各站点行程数量如下：
宾夕法尼亚州：16344次行程
华盛顿州：22974次行程
佛罗里达州：67163次行程
印第安纳州：21300次行程
北卡罗来纳州：49313次行程
纽约州：69543次行程

第二阶段：
三类驾驶时段的数据：
1. 巴特尔识别出的750起与超速相关的安全关键事件
2. 巴特尔从现有平衡基准组与额外基准组中选定的1532个基准时段
3. 巴特尔选定的260个前置视频行程片段
安全关键事件与基准时段均附带事件详情。上述两类分组的事件前60秒数据均包含时序数据与Navteq地图数据。前置视频行程片段的全时段数据均包含时序数据、Navteq道路数据及前置视频。
为保护参与者的个人可识别信息（Personally Identifiable Information，简称PII），本研究采用了起点与终点缓冲区机制。与缓冲区重叠的时段将不提供位置信息与前置视频（如适用）。撞车事故的位置信息及撞车事件的前置视频均不予公开。

数据规格
本数据集包含的变量可参阅随附的数据字典文件：
- 关键事件与基准时段的视频缩减数据：请参阅Event_Table_Data_Dictionary.pdf
- 驾驶员ID与驾驶员人口统计学数据：请参阅DriverDemographic_DataDictionary.pdf
- 行程摘要数据：请参阅Trip SummaryDataDictionary.pdf
- 行程时序采样数据：请参阅TimeSeriesData Dictionary.pdf
时序数据以1赫兹（1Hz）的频率进行规整。缺失纬度或经度信息的文件已被排除。

额外分析与超速数据库开发
本数据集包含为《SHRP2超速数据分析》研究开发的数据缩减产物，以及本次《SHRP2超速数据：额外分析与超速数据库开发》研究中生成的全新数据。
《SHRP2超速数据分析》研究依托SHRP2自然驾驶研究中六个数据采集站点的样本数据，对超速问题展开了分析。各站点的时序数据经处理后，可识别出驾驶员拥有超速空间的自由流时段（Free-Flow Episodes，简称FFEs）与超速时段（Speeding Episodes，简称SEs）。本研究共开发了三类用于描述对应行程、自由流时段及超速时段内驾驶行为的数据缩减产物：（1）行程级数据缩减；（2）自由流时段（FFE）数据缩减；（3）超速时段（SE）数据缩减。上述数据集的生成与分析细节可参阅该研究的《结果报告》（Richard等，2018）与《方法报告》（Brown & Richard，2018）。
本次研究为解答其中一项研究问题（即常规交通场景下的超速与激进驾驶行为具有何种特征？），开发了两项全新的数据缩减产物。这些缩减产物可为研究人员提供数据，用于分析超速状态下的激进变道与跟车过近行为。本数据缩减产物及额外信息可参阅下方“文件”标签中的《SHRP2超速数据：额外分析与超速数据库开发》压缩文件夹。

搜集汇总

背景与挑战

背景概述

该数据集是基于SHRP2自然驾驶研究的超速行为分析，包含两个阶段的研究数据。第一阶段探索了超速行为，第二阶段则利用视频数据进一步研究超速和攻击性驾驶行为，涉及多个地区的驾驶行程和安全事件数据。