Design of the In-Vehicle Driving Behavior and Crash Risk Study (2011)

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 to Be CollectedDriver Demographics and Vehicle Inventory Basic demographic data will be gathered from each primary driver. The list currently includes the following: • Gender; • Date of birth; • Ethnicity; • Race; • Country of birth; • Education level; • Marital status; • Household makeup; • Household ownership; • Working status; • Vocation; • Household income; • Household population (i.e., number of people living at par- ticipant’s residence); • Household age categories; • Number of vehicles (i.e., number of vehicles residing at the participant’s residence); • Zip code; • Years of residence (i.e., number of years of residence in cur- rent neighborhood); • Vehicular travel (i.e., estimated number of miles driven last year by the participant); • Business use (i.e., is the instrumented vehicle used for busi- ness purposes?); • Business purpose (i.e., if the instrumented vehicle is used for business purposes, what are those business purposes?); and • Licensure age (i.e., at what age did the participant receive his or her first driver’s license?). In addition, the instrumented vehicle will be inventoried to record its basic facts as well as additional options and features. This inventory will include at least the following:13• VIN (vehicle identification number, nonpersonally identi- fying digits only); • Make; • Model; • Year; • Style/trim level (e.g., LX, EX, EX-L, or Touring for the 2010 Honda Odyssey); • Body style (e.g., coupe, sedan, or wagon); • Color; • Safety features (e.g., antilock brake system [ABS]; electronic stability control [ESC]; front, side, curtain airbags; forward collision warning [FCW]; lane departure warning [LDW]; adaptive cruise control [ACC]); and • Infotainment features (e.g., integrated navigation or info- tainment systems). Driver Assessment In a study such as this, it is desirable to measure certain driving- related functional capabilities and limitations of the partici- pants. These, singly or combined in some fashion, may help to explain some of the variability in the driving or crash-related data observed. If so, such factors may lead to the development of countermeasures targeting such individual differences or impairments. As part of the development of the driver testing suite, a review of the relevant literature was conducted to help devise the assessment plan. In addition, a blue-ribbon panel of experts thoroughly reviewed all aspects of the plan and provided invalu- able, iterative feedback. All of their concerns were addressed in some fashion, and all members of the panel ultimately indi- cated their approval of the plan that is presented herein. On the basis of the exercise described, an approach was defined wherein each participant will be assessed on several dimensions thought to be relevant to driving and poten- tially predictive of some portion of driving behaviors, prob- lems, or crash-related events (e.g., crashes and near crashes). A

14systematic approach was undertaken to identify the most rele- vant dimensions and essential criteria in the development of the resulting driver assessment testing plan. The sleep questionnaire was constructed in close coordination with the Harvard Medical School Division of Sleep Medicine. The selected relevant dimensions were executive function and cognition; visual perception; various visual–cognitive, physical, and psychomotor abilities; personality factors; sleep- related factors; medicines and medical conditions; driving knowledge; and history. Executive function (EF) broadly encompasses a set of cognitive skills that are responsible for the planning, initiation, sequencing, and monitoring of com- plex goal-directed behavior (Royall et al. 2002). The criteria used in selecting the testing instruments to measure each of these dimensions were comprehensiveness, evidence of pre- dictive value, feasibility of administration, uniqueness, per- sistence, and feasibility of replication. The driver assessments are intended to be administered in a 2- to 3-hour period that will, for the most part, occur simul- taneously with vehicle instrumentation. A pilot test of the suite of driver assessments was conducted and, on the basis of the outcome of that pilot test, it is believed that they can be administered within that time frame. The current driver test- ing assessments are listed below. Some instruments will be administered by trained person- nel; other instruments include online questionnaires that canbe filled in at the assessment site or later (e.g., within a week of installation). The first participant compensation payment will be made after all driver assessment activities, including filling in all questionnaires, have been completed. The assessments to be implemented are listed in Tables 3.1 to 3.7.In Person/ Estimated Measurement Construct Description Online Time (min) Comments Spatial relationships In person 3 ▪ Central vision and processing speed In person 5 ▪ Divided/selective attention Divided attention In person 6 In person Table 3.2. Visual-Cognitive Assessments Included in DrivingHealth Inventory software Motor-Free Visual Perception Test (MVPT) Visual Closure Subtest Useful Field of View (UFOV)—Part 2 only Trail-Making Test (A & B) Rapid Pace Walk (discussed in Table 3.4 below but also included here, since it is recorded within the DrivingHealth Inventory software) In Person/ Estimated Measurement Construct Description Online Time (min) Comments ▪ Executive function/working memory In person 15 ▪ Reaction time Dementia In person 2 Table 3.3. Cognitive and Psychomotor Assessments Administered to all participants Connors’ Continuous Performance Test II (CPT II), Ver. 5.1 Clock-Drawing Test: ▪ Draw a clock ▪ Put in all the numbers ▪ Set the hands at 10 past 11Measurement In Person/ Estimated Construct Description Online Time (min) High-light In person 5contrast sensitivity Low-light 5contrast sensitivity Near static 5 acuity Far static 5 acuity Depth 5 perception Color vision 5 Peripheral 5 vision Table 3.1. Visual Perception Assessments Optec 6500P (Figure 3.1)

15Measurement In Person/ Estimated Construct Description Online Time (min) Lower limb In person 1strength/ mobility Upper body In person 2 strength Table 3.4. Physical Assessments Rapid Pace Walk Jamar Hand DynamometerIn Person/ Estimated Measurement Construct Description Online Time (min) Sleep quality Online 10 Potentially driver-impairing medical conditions Online 20 and medications Table 3.5. Health-Related Assessments Harvard sleep questionnaire Custom comprehensive questionnaire based on material from the American Medical Association (AMA) and National Highway Traffic Safety Administration (NHTSA) In Person/ Estimated Measurement Construct Description Online Time (min) Risk-taking behavior Online 2 Risk perception Online 2 Attention deficit hyperactivity disorder (ADHD) Online 2 Driver style/behavior Online 5 Thrill/adventure seeking Online 5 Table 3.6. Psychological Assessments Cox Assessment of Risk Driving Scale (CARDS) combined with DeJoy Risk Perception Questionnaire CARDS items combined with DeJoy Risk Perception Questionnaire Barkley’s Adult ADHD Quick Screen Manchester Driver Behavior Questionnaire (DBQ) Sensation Seeking Scale (SSS) Measurement Construct Description In Person/Online Estimated Time (min) Driving knowledge Online 10 Driving history Online 5 Table 3.7. Driving Knowledge/History Assessments Custom questionnaire based on several state licensing practice tests Custom questionnaireData Acquisition System Variables The DAS comprises three primary units: the head unit, the main unit, and the front radar assembly as illustrated in Fig- ure 3.2. Cabling will provide hard-wired connections between the head unit and main unit as well as a power source. Addi- tionally, cabling will connect to the onboard diagnostic port(OBD II) so that specific vehicle network data can be obtained. Turn signals are collected from network data as available, or a cable connection will be established to capture turn-signal usage data. Note that the head unit is actually composed of two subassemblies: the rear camera and cellular antenna (General Packet Radio Service, GPRS), both located on the rear pack- age shelf (or similar location in SUVs and minivans). The head unit (Figure 3.3) assembly holds three cameras that will capture video images as illustrated in Figure 3.4. These image components include the forward roadway (upper left), driver face (upper right), and the pedals and instrument clus- ter interactions (lower left). The area behind and to the right of the subject vehicle (lower right) will be captured by a sep- arate camera mounted on the rear of the cabin. Note that the face image is oriented at a 90-degree angle from the other images to maximize the efficiency of pixel allocation among all four images. That image and all images will be correctly oriented before any type of video analysis. Also, another camera, located in the head unit, will periodically take a still

16Front Turn Signals Radar Interface Box Radar Unit PWR Bluetooth PWR DAS Main Unit OBD Connector Head Unit Sub-Head Unit GPRS Antenna Rear Camera Figure 3.2. DAS schematic.Figure 3.1. Optec 6500P all-in-one vision test apparatus.image of the cabin. These still images will be used to deter- mine the presence of passengers, and they will be irretrievably blurred to protect the anonymity of unconsented passengers. This assembly will be mounted to the right and rear of a par- ticipant’s rearview mirror. The forward camera will capture a color image, and the remaining cameras will be low-cost, small-form-factor monochromatic cameras that are sensitive to infrared (IR) illumination (for low-light video capture). The forward camera will collect color video to provide more comprehensive information about the forward or driver’s view. Furthermore, for those researchers interested in traffic signal state detection, the color camera will provide the ability to conduct such research through a post hoc machine appli- cation or manual data analysis. Additionally, the alcohol sen-sor, incident button, illuminance sensor, inertial acceleration, and gyroscopic sensors will be incorporated into the head unit assembly. In determining which sensors could be the most valuable, an extensive review of the research questions was conducted. The outcome of that prioritization led to the chosen sensors. The selected sensors reflect the best combination of desir- able data that can be obtained within the cost and engineering constraints of the project and that meet the highest-priority research questions. The main unit (Figure 3.5) is host to the computer functions of coordinating sensor nodes, communications (internally and via the cellular capability), and data storage on the hard drive (HD). It will have the capability of storing data on an onboard solid state HD for 4 to 6 months of typical driving. It will fea- ture continuous asynchronous data collection. That is, each sensor or variable will be recorded at its native frequency and time-stamped with a real-time clock, without regard to the fre- quency at which any other data are being recorded. This is the most efficient approach to dealing with storage limitations. Vehicle network information is expected to be obtained from most of the vehicles for driver control interactions. For some makes and models, additional advanced technology variables are expected to be captured, such as the ABS, ESC, brake assist, traction control, etc. This feature requires OEM cooperation as described previously. Accelerometers will nominally collect X, Y, Z acceleration at 10 Hz, but these data will be continu- ously buffered at 500 Hz and saved at this higher rate under certain circumstances to provide higher-resolution accelera- tion information during a crash event. Yaw rate gyro will also be collected continuously at 10 Hz.

17Figure 3.3. Head unit assembly prototype.Figure 3.5. Main unit prototype.Figure 3.4. Composite snapshot of four continuous video camera views.The GPRS cellular antennae will be mounted in the rear of the cabin. Wide field of view (FOV) forward radar capable of assessing oncoming traffic will be small, lightweight, and mounted to the front license plate holder (Figure 3.6). Radar data will be transmitted via Bluetooth wireless technology to the main unit and will capture data concerning the relative position and speed of multiple objects. Use of Bluetooth for this purpose alleviates the need of running cables through the vehicle’s firewall to the main unit, thereby decreasing risk to the participant and saving valuable installation time and real costs associated with permanently altering a participant’s vehicle. This is important because each S07 site must maintain an installation throughput rate of two vehicles per bay per day (i.e., where the largest sites have three installation bays and the smallest have a single installation bay). DAS sensors and associated variables are listed in Table 3.8.Figure 3.6. Forward radar assembly prototype.Machine Vision Applications Several machine–vision-based applications will be incorpo- rated into the DAS, including a lane tracker, head position monitor, and driver identification software. There are crite- ria for what to include as resident software on the DAS, as opposed to that which can be applied to stored data post hoc. The video data are being stored in a compressed format, because there are insufficient resources to store it in its native resolution. However, several of the machine–vision-based applications require the full resolution video to perform as expected, so these must operate on the native video coming directly from the cameras before the information is stored on the DAS hard drive. On the other hand, every such applica- tion running in real time consumes limited storage and pro-

18Location Sensor Data to Be Collected Head unit Main unit Main unit Head unit Radar unit Main unit Head unit Head unit Head unit Head unit Main unit Table 3.8. Data Acquisition System Variables Multiple cameras/video Accelerometer data Rate sensors GPS Forward radar Cell phone module Illuminance sensor Passive alcohol sensor Incident push button Audio Turn signals (from vehicle network data or directly from the signals themselves) Vehicle network data Video images of the forward view, center stack view, rear and passenger side view, and driver face view, information for machine vision (MV) processes (including lane tracking and eyes forward data), as well as periodic, irrevocably blurred still photographs of the cabin interior to capture passenger presence In 3 axes: Forward/reverse (x) Right/left (y) Down/up (z) Yaw rate Latitude, longitude, elevation, time, velocity Object ID, range, and range rate Health checks, location notification, collision notification, and remote software upgrades Ambient lighting levels Presence of alcohol within the vehicle cabin In the event of an unusual or interesting traffic safety-related event, allows participant to open an audio recording channel for 30 seconds; also “flags” the data stream for ease of location during data analysis Available only in concert with the incident push button as noted above Turn signal actuation, which distinguishes between left and right indicated turns Where available, the use of the accelerator, brakes, ABS, gear position, steering wheel angle, speed, horn, seat belt information, airbag deployment, and many other such variablescessing resources. Therefore the design must balance the need that certain applications have for real-time, onboard process- ing with the storage and processing limitations of the DAS. The VTTI Road Scout (VRS) application is designed to track lane markings in real time on the DAS. Having this appli- cation on the DAS provides the advantage of operating on uncompressed video. The VRS has the ability to determine location of lane lines, horizontal curvature of the road, and angular offset of the vehicle within the lane. The VRS can determine if the lane lines are single, double, solid, or dashed. Testing has shown that the lane-tracking algorithm has a high degree of accuracy when lane markings are clearly present and visible; however, the lane tracking functionality is unable to gather sufficient roadway data in conditions where snow or other occlusions are present in the roadway. The data available from the MV lane tracker will be useful in answer- ing many questions about road departure, because it is antic- ipated that many road departures and unintentional lane departures will be captured during the SHRP 2 NDS.The VTTI Mask Head Tracker also operates on the DAS so that it can operate on uncompressed video (Figure 3.7). It is capable of identifying and distinguishing between a few gen- eral glance locations (e.g., forward roadway, mirrors, center stack) using software designed to find and determine charac- teristics of a person’s face in an image, and track those char- acteristics through subsequent images collected with the face view camera. The mask generates a three-dimensional repre- sentation of a person’s face using triangular surfaces to define the shape. The software and system operation will function in real time on the DAS using the raw video before compression. A face recognition software solution is being sought that will permit researchers to automatically determine if a driver is a consented participant. Systems of this type are too processing- intensive to operate in real time, so this type of processing and analysis will be done on a post hoc basis. Driver identification will rely on a biometric software application to provide auto- mated face recognition of drivers on the basis of their unique facial characteristics. Face recognition software would substan-

19Figure 3.7. VTTI Mask Head Tracker: Calibrated eyes forward (left); calibrated eyes on speedometer (right).tially reduce reliance on a human data reductionist to open each trip file to perform visual verification of the participant. Crash Investigation Data from crash investigations provide an important comple- ment and extension of the naturalistic driving data that will be collected as part of the NDS. While the data captured by the instrumented vehicle will be extensive, it is expected that they will not be a complete record of every detail of a crash, so the methodology recommended here is designed to collect the additional data needed in a way that is both feasible and effective. Data from the crash investigations should signifi-cantly enhance understanding of the actions, conditions, and behaviors that led to the crash. In addition, the comparison of crash investigation data, police accident reports (PARs), and DAS data will provide interesting insights into the reliability of these different sources of crash information. Given that more than 1,000 crashes of all severity levels (as well as perhaps an order of magnitude more of number of near crashes) are expected during this study (see Table 3.9), not all crashes recorded during the study are expected to be investi- gated because of several constraints, including cost and, since many crashes are expected to be minor, the lack of crash site information. Crash investigations will only be carried out for crashes that meet certain criteria of interest. Such criteria mayBased on the 100-Car Study, Modified Based on Crash Rates Crash/Incident Severity 100-Car Study by Fatality Rate from GESa and NHTSb Police reported 624 363 230 Nonreported, reportable 975 566 360 Nonpolice reported: low-g contact or tire strike 1,599 929 590 Total crashes 3,198 1,859 1,180 Near crashes 29,679 17,247 10,952 a General Estimates System. b National Household Travel Survey, 2001 (FHWA 2010). Table 3.9. Crashes and Near Crashes in the SHRP 2 Naturalistic Driving Study Estimated by Three Methods (Based on 1,950 DASs for 2 Years)

20be altered as the study proceeds, but they could include such factors as the following: • Severity (e.g., airbag(s) deployed or injuries sustained); • Crash type (e.g., intersection-related or lane change/ ran-off-road); • Driver age (e.g., teen driver or older driver); • Land use (e.g., rural versus urban); and • Advanced technology vehicle (e.g., equipped with crash warning/avoidance technology). The rates based on the 100-Car Study (Dingus et al. 2006) (see the first data column in Table 3.9) simply and directly extrapolate the crash rates observed in the 100-Car Study to the size and scope of the current study. These estimates were then modified on the basis of the ratio of the relatively high fatality rates for Washington, D.C. (the site of the 100-Car Study) and that of the United States overall (see the second data column in Table 3.9). Fatal crashes were selected, since General Estimates System (GES) crash estimates are not avail- able for a particular locality, such as the Washington, D.C., and northern Virginia area. According to Fatality Analysis Reporting System data, Washington, D.C., had a fatality rate of 29.15 per 100,000 registered vehicles in 2003 compared with the latest available national rate of 16.05 per 100,000 reg- istered vehicles in 2006 (NHTSA 2010; NHTSA 2007). This analysis assumes that the relative traffic fatality rates between Washington, D.C., and that of the United States overall can also be used to roughly approximate the relative crash and near-crash rates. The numbers in the third data column in Table 3.9 are derived from GES and 2001 National Household Travel Survey (FHWA 2010) estimates. Near crashes have been operationally defined by VTTI researchers (Dingus et al. 2006) as any circumstance that requires a rapid, evasive maneuver by the subject vehicle, or any other vehicle, pedestrian, cyclist, or animal to avoid a crash. A rapid, evasive maneuver is defined as steering, brak- ing, accelerating, or any combination of control inputs that approaches the limits of the vehicle’s capabilities. However, there is nothing prohibiting other researchers from defining the concept as they see fit. In most cases, S06 personnel will be notified of a crash by the DAS via cellular communication channels. Included in the communications process will be key details about the crash, including a snippet of video covering the time just before and immediately after the crash. S06 personnel will then assess those data against the crash severity criteria bulleted above to determine if the crash event warrants further investigation. If the crash is selected for further investigation, the S07 contractor will make an attempt to gather or retrieve the fol- lowing information: • DAS data. • PAR (as available from participant or public records).• Participant interview (using an instrument provided by the S06 contractor)—as soon as feasible postcrash to determine the following:  Predrive factors: ▪ Recent sleep patterns; ▪ Fatigue levels; ▪ Emotional states; and ▪ Stress levels.  Driving factors: ▪ Weather; ▪ Traffic; and ▪ Obstructions.  Crash factors. • Aerial view (as available, for example, from Google Earth or Google Street View). • Vehicle photos (can be taken by data retrieval technicians)— front, back, sides. • 37 Crashes categorization (Najm et al. 2007). Considering data privacy and IRB issues, the only people to be interviewed in all crash investigations will be the consented drivers (and possibly other consented passengers) participat- ing in the SHRP 2 NDS. Note that all the crash data noted above can be gathered remotely or with already-existing personnel. A detailed list of data elements to be collected has been selected to be consistent with common data elements in the National Center for Statistics and Analysis national crash data, either by adopting the structure or by structuring indi- vidual data elements so that they can be mapped into the national data. Investigations will include documentation and data collection as related to precrash driver assessment, inter- views, the crash site, and vehicle examination (as available). The type of precrash and postcrash assessment information to be collected will be similar to the recent National Motor Vehicle Crash Causation Survey (NMVCCS) conducted by NHTSA (2008). On the basis of the criteria of severity, type (e.g., run-off- road or intersection related), or some condition of uniqueness, the S06 contractor will notify the S07 contractor to dispatch an experienced investigator to visit the site within 48 hours (but not while police/emergency medical service [EMS] personnel are actively working the crash scene) to produce or retrieve the following additional information: • A detailed description of the crash etiology; • Crash site documentation and description (using software such as Easy Street Draw); • Crash-site photographs showing the approach to point of impact for each involved vehicle and looking back from the point of impact for each vehicle; and • Photos of physical evidence such as skid marks, gouges in the roadway or median, and impact points. Crash-site investigation is expected to be conducted by expe- rienced crash-site/scene investigators.