National Academies Press: OpenBook
« Previous: Chapter 2 - Identifying Final Lane Departure Research Questions and Relevant Factors
Page 22
Suggested Citation:"Chapter 3 - Data Sets Used." National Academies of Sciences, Engineering, and Medicine. 2011. Evaluation of Data Needs, Crash Surrogates, and Analysis Methods to Address Lane Departure Research Questions Using Naturalistic Driving Study Data. Washington, DC: The National Academies Press. doi: 10.17226/22848.
×
Page 22
Page 23
Suggested Citation:"Chapter 3 - Data Sets Used." National Academies of Sciences, Engineering, and Medicine. 2011. Evaluation of Data Needs, Crash Surrogates, and Analysis Methods to Address Lane Departure Research Questions Using Naturalistic Driving Study Data. Washington, DC: The National Academies Press. doi: 10.17226/22848.
×
Page 23
Page 24
Suggested Citation:"Chapter 3 - Data Sets Used." National Academies of Sciences, Engineering, and Medicine. 2011. Evaluation of Data Needs, Crash Surrogates, and Analysis Methods to Address Lane Departure Research Questions Using Naturalistic Driving Study Data. Washington, DC: The National Academies Press. doi: 10.17226/22848.
×
Page 24
Page 25
Suggested Citation:"Chapter 3 - Data Sets Used." National Academies of Sciences, Engineering, and Medicine. 2011. Evaluation of Data Needs, Crash Surrogates, and Analysis Methods to Address Lane Departure Research Questions Using Naturalistic Driving Study Data. Washington, DC: The National Academies Press. doi: 10.17226/22848.
×
Page 25
Page 26
Suggested Citation:"Chapter 3 - Data Sets Used." National Academies of Sciences, Engineering, and Medicine. 2011. Evaluation of Data Needs, Crash Surrogates, and Analysis Methods to Address Lane Departure Research Questions Using Naturalistic Driving Study Data. Washington, DC: The National Academies Press. doi: 10.17226/22848.
×
Page 26

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

C H A P T E R 3 Data Sets UsedThe following sections describe the data sets used throughout this report for the various analyses. Appendix A provides a detailed description the variables included in the UMTRI data set, as well as a description of how a number of variables were extracted. Appendix B provides a description of variables avail- able in the VTTI data set. University of Michigan Transportation Research Institute Field Operational Test In-Vehicle Data Several field operational tests were conducted by UMTRI, including a road departure crash warning (RDCW) system. The system involved mounting instrumentation packages on 11 vehicles (of the same make and model). In each of the studies, vehicles were instrumented with a variety of sensing systems, including a forward video and a driver’s face video, forward radar and side radar, and a global positioning system (GPS). The RDCW system also used a camera to record visual features that delineated the lane and road edges, and radars that monitored the lane edge. The main advantage of this data set was that the researchers were able to archive all the data so the database could be searched and specific data extracted. The RDCW included a lane departure warning (LDW) sys- tem and a curve speed warning (CSW) system. The LDW sys- tem used a forward-looking monochrome camera to identify visual features near the lane edge. The image position of visual features and algorithms were used to calculate lane width, vehicle position within the lane, and relative motion within the lane. Other sensors, such as GPS, vehicle speed, brake posi- tion, and forward and side radars, were used to increase the accuracy and reliability of determining the lane position. The CSW system processed road geometry to estimate curvature and then, using vehicle speed and acceleration, computed a vehicle’s most likely path and risk of leaving the curve (LeBlanc et al., 2006).22The RDCW system alerted the drivers when they drifted from their lanes or went too fast to safely negotiate a curve. The system was tested over 10 months with 78 drivers who were evenly split by gender and age. Data were collected for a one- week period prior to activation of the system for each driver. During the first week of driving, the RDCW was functional and recording data just as if the system were operational, but alerts were not provided to the driver. As a result, the first week of data collection reflected naïve driving with no in-vehicle warn- ing system alerts. The RDCW system included six levels of alerts that would have indicated to the driver that he or she was about to leave his or her lane or was traveling too fast on a curve. The alerts included right- and left-lane departure cautionary alerts, right- and left-lane departure imminent alerts, and cautionary CSW and imminent CSW alerts. A seventh designation was used to indicate that a vehicle was negotiating a curve, but this did not include any alert. The research team requested data from the road departure crash warning (RDCW) field operation test (FOT) for the one-week period prior to activation of the RDCW system for all instances when one of the six alerts was recorded. Data for all rural roadways were requested. Rural data are defined in the UMTRI FOT as data from a location with a population less than 50,000 persons. Data for instances when alerts were recorded were used as starting points to identify potential lane departures. This infor- mation was used as described in Chapter 5 to develop thresh- olds for crash surrogate events. The team also requested data on regular driving that did not involve any alert. UMTRI provided data in the form of a database, as well as of forward imagery. Data were provided for 44 different driv- ers and were divided by alert type. The research team received over 2,000 alerts (1,506,525 rows of data). Each alert included approximately 600 rows of data at 10 Hz (one row represents 0.1 s). Data were provided for 30 s before the alert was recorded and approximately 30 s after. Instances of a vehicle negotiating

23a curve were also provided as samples of regular driving and included approximately 60 s of data. The database contained a number of data fields (columns) with data from the instru- mentation system, such as lateral acceleration and forward speed. The data fields included are described in Appendix A. The data corresponding to each alert are referred to in the fol- lowing sections as “vehicle trace,” as shown in Figure 3.1.Image source: Esri. © 2010 i-cubed. Data source: UMTRI RDCW data set. Figure 3.1. Example of vehicle trace.Forward imagery was provided for each vehicle trace. Images were provided at 2 Hz (2 per s or 1 image per 5 rows of vehicle trace data) during times when an alert was not recorded. Data were provided at 10 Hz (10 per s or 1 image per row of vehicle trace data) for the 4 s before and 4 s after an alert was recorded. Imagery was provided as compressed JPEG images. Images had to be lined up with the corresponding rows of data. An example forward view is shown in Figure 3.2.Source: UMTRI RDCW data set. Figure 3.2. Example of forward video view.Hereafter, this data set will be referred to as the “UMTRI naturalistic driving study data set” or “UMTRI data set.” Data included several roadway types, including rural free- way, rural freeway ramp, rural expressway, rural four-lane, rural two-lane paved, and rural two-lane gravel. Since a large amount of data was provided, the team focused on two-lane roadways to meet project goals and deadlines. The lane track- ing system did not perform well on unpaved roadways, so the study further focused on two-lane, paved roadways.The data set included a database file with the following variables: • Driver number; • Trip number; • Alert time; • Time; • Alert ID; • Alert type; • Age; • Gender; • Curve (present or not); • Right and left boundary types (type of lane marking); • Latitude and longitude (used to establish vehicle position); • Heading; • Available maneuvering room, right and left (distance to left and right measured by radar); • Brake (brake engaged or not); • Engaged (cruise control engaged or not); • Lane offset (vehicle offset from lane center); • Lane offset confidence; • Lane width; • Track width (width of vehicle; consistent, since same vehicle model was used); • Speed; • Lateral speed; • Side and forward acceleration; • Roll, roll rate, pitch rate, yaw rate; • Solar angle; • Wiper (e.g., off, low); • Headlamp (off, on, parking, high); • Road class (roadway type: unknown, limited use, major surface, minor surface, local); • Curve advisory speed, if present; • Posted speed limit; • Curve radius; • Distance to curve point of intersection; • Annual average daily traffic (AADT); and • Number of lanes. A variable was also included for the widths of the left and right shoulders. However, all shoulders were indicated in the data as being 5 m, so it was assumed that this variable was incorrect. Shoulder width was then measured using the for- ward imagery instead. Michigan Geographic Framework and Sufficiency Report The Michigan Geographic Framework (MGF) is the digital base map for the state of Michigan. Public roads are one of the

24many features maintained as part of this framework. Roadway attributes include linear referencing descriptors, road name, address ranges, functional class, and legal ownership. The loca- tion of most roadway-based data is described using the MGF linear references. The Michigan Department of Transportation (MDOT) suf- ficiency log is an annual report created by the Office of Trans- portation Planning for the trunk line (state-maintained) highways. The sufficiency report includes a broad range of roadway attributes. Sample attributes include MGF-based lin- ear references, road type, surface and shoulder type, surface and shoulder width, number of through and turn lanes, traffic volume, and various sufficiency-based ratings. The databases were obtained as part of another project at CTRE. MDOT confirmed through e-mail that the researchers are allowed to use the data set for the research described in this report. Transportation Crash Master The Transportation Crash Master is an extract of MDOT’s crash report information system (CRIS) database. It contains general information about crashes on all public roads, includ- ing attributes for up to three vehicle units. Crash location is provided through both MGF-based linear references and derived geographic coordinates. The database was obtained as part of another project at CTRE. MDOT confirmed through e-mail that the researchers can use the data set for the research described in this report. One stipulation of using the data is that crashes cannot be shown in a map in any document or presentation. Data were available for 2000 to 2006. Aerial Imagery Aerial imagery from Esri Corporation was also used. The team had access to a program in Esri’s geographic informa- tion system (GIS), ArcMap, which interactively brings up aer- ial imagery for a location where other spatial data are already available. The imagery is hosted by Esri and available online (see http://resources.esri.com/arcgisonlineservices). The data come from a variety of sources, including U.S. Geological Survey (USGS) Digital Orthophoto Quarter Quadrangles (DOQQs). Figures 3.3 and 3.4 show vehicle traces overlaid with aerial imagery in ArcMap. Google Earth was also used as a source for aerial imagery.Image source: Esri. © 2010 i-cubed. Vehicle trace source: UMTRI RDCW data set. Figure 3.3. Vehicle traces overlaid with aerial imagery in ArcMap.Image source: Esri. © 2010 i-cubed. Vehicle trace source: UMTRI RDCW data set. Figure 3.4. Single-vehicle trace overlaid with aerial imagery in ArcMap.VTTI Naturalistic Driving Study from Data Request The team requested all rural lane departure crashes, near crashes, and incidents from the VTTI 100-car naturalisticdriving study that occurred in open country or on the inter- state. A sample of nonevent data was also requested. The team requested reduced data from events; raw data from the instru- mented vehicle’s data acquisition system (DAS); and forward, side, and back video. The team also requested samples from nonevent driving. The team received a total of 33 crashes or near crashes on roadways classified as open country or interstate run-off-road. No incident or nonevent data were

25received. It is unknown if the data represented all run-off-road crashes and near crashes. A number of variables were requested for the raw data. The raw data received included the following variables: • Trip time; • Trip time elapsed; • Event start and end time; • Side and forward acceleration; • Speed; • Throttle (throttle position); • Available maneuvering room forward and rear (included range, rate, and azimuth); • Brake (on or off); • Turn signal state (off, right, left); • Wiper state; • Light condition; • Urban/rural setting; and • Weather condition. A variable with latitude and longitude was provided, but its data fields were invalid and, as a result, vehicle data could not be spatially located. Other data from the DAS that the team believed were available and requested but were not received included the following: • Vehicle position within lane and lane width from lane tracking; • Curve advisory speed and radius; • Volume; • Posted speed limit; • Shoulder width; and • Shoulder type. A video clip was provided for each event that included approximately 45 s of data. Images showing a forward, rear, and right-side outside view were provided. It is assumed that the forward view provides as much coverage (distance outside the front of the vehicle viewed in the image) as was available in the original data set. It is also assumed that the image reso- lution is the same as was available in the original VTTI data set and was not degraded. VTTI Naturalistic Driving Study from the Internet VTTI recently made some data from their 100-car study pub- licly available on a data distribution website (http://forums .vtti.vt.edu/index.php?/files/category/3-100-car-data/). The following data were obtained: • Continuous data: Contains data around each crash or near crash at 10 Hz.• Event eyeglance data: Has recording of participant eye- glance location for each crash or near crash. • Event video reduction data set: Contains reduced data from the video for crashes and near crashes and has information such as event nature, event type, and precipitating event. • Event narrative: Has narrative for each crash/near crash. This data set had data for each of the 33 crashes/near crashes provided by VTTI. Some events were also available for eight additional crashes/near crashes that may be used in the analy- sis of lane departures. The data set provided some additional information, such as throttle position, lane markings, and dis- tance to objects left and right. SHRP 2 Full-Scale Instrumented Vehicle Study SHRP 2 is in the process of implementing a large field study of instrumented vehicles driven by naïve drivers. The naturalistic driving study will instrument approximately 2,000 vehicles in six states (Indiana, Pennsylvania, Florida, New York, North Carolina, and Washington) (Campbell, 2009). The instrumen- tation will be left in place for 12 or 24 months. Roadway data will also be obtained through several sources. Roadway databases from study states will be obtained. SHRP 2 also plans to collect a limited amount of high-resolution road- way data using deployed mobile mapping units. A description of the data that are expected to be collected and that are relevant to answering lane departure research questions is provided in Chapter 4. Summary of Terms Used to Describe Data • Continuous data: This type of data is described more in Chapter 6. Continuous data are naturalistic driving study data reported at the resolution at which they were collected. For instance, each row represents one observation of vehi- cle driving for 0.1 s (10 Hz). This is also referred to in some studies as “time series data.” • Dynamic: Dynamic variables are those variables that change in the short term. These include vehicle kinematic variables such as speed, acceleration, or position. In reality, most roadway variables would not change over the course of the study, but since the roadway a driver is currently traversing will change in the short term, roadway characteristics are considered to be dynamic variables. All environmental vari- ables are also considered to be dynamic. • Event: This is described more in Chapter 8. An event is an interval of time centering on a situation of interest, such as a lane departure. For instance, an event may consist of 30 s of data before and after a lane departure occurs.

26• Frame: A frame is one row of data from the continuous data. At 10 Hz this represents a 0.1-s interval. • Incident: An incident is an occurrence of a situation of interest, such as a lane departure. • Static: Static is a way to describe variables that do not change over the course of the study, such as driver gender, and results from preinstrumentation driver surveys. Age is likely to change during the course of the study, but for all intentsand purposes age can be considered a static variable. Vehi- cle static variables include vehicle type, track width, and cen- ter of gravity. • Vehicle trace: This is used to describe the intervals of data provided in the UMTRI data set. Each vehicle trace consisted of approximately 60 s of data at 10 Hz (around 600 rows of data), which could be spatially located.

Next: Chapter 4 - Roadway, Driver, Environmental, and Vehicle Data Needs and Limitations to Address Lane Departures Using Naturalistic Driving Study Data »
Evaluation of Data Needs, Crash Surrogates, and Analysis Methods to Address Lane Departure Research Questions Using Naturalistic Driving Study Data Get This Book
×
 Evaluation of Data Needs, Crash Surrogates, and Analysis Methods to Address Lane Departure Research Questions Using Naturalistic Driving Study Data
MyNAP members save 10% online.
Login or Register to save!
Download Free PDF

TRB’s second Strategic Highway Research Program (SHRP 2) Report S2-S01E-RW-1: Evaluation of Data Needs, Crash Surrogates, and Analysis Methods to Address Lane Departure Research Questions Using Naturalistic Driving Study Data examines the statistical relationship between surrogate measures of collisions (conflicts, critical incidents, near collisions, or roadside encroachments) and actual collisions.

The primary objective of the work described in this report, as well as other projects conducted under the title, Development of Analysis Methods Using Recent Data, was to investigate the feasibility of using naturalistic driving study data to increase the understanding of lane departure crashes.

This publication is available only in electronic format.

READ FREE ONLINE

  1. ×

    Welcome to OpenBook!

    You're looking at OpenBook, NAP.edu's online reading room since 1999. Based on feedback from you, our users, we've made some improvements that make it easier than ever to read thousands of publications on our website.

    Do you want to take a quick tour of the OpenBook's features?

    No Thanks Take a Tour »
  2. ×

    Show this book's table of contents, where you can jump to any chapter by name.

    « Back Next »
  3. ×

    ...or use these buttons to go back to the previous chapter or skip to the next one.

    « Back Next »
  4. ×

    Jump up to the previous page or down to the next one. Also, you can type in a page number and press Enter to go directly to that page in the book.

    « Back Next »
  5. ×

    To search the entire text of this book, type in your search term here and press Enter.

    « Back Next »
  6. ×

    Share a link to this book page on your preferred social network or via email.

    « Back Next »
  7. ×

    View our suggested citation for this chapter.

    « Back Next »
  8. ×

    Ready to take your reading offline? Click here to buy this book in print or download it as a free PDF, if available.

    « Back Next »
Stay Connected!