Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 74
74 Besides these existing warning systems, potentially bene- time and location stamp in the data set, naturalistic data can ficial warning systems not yet tested might be effective in be highly efficient in recognizing the relationship between reducing safety-related events; for example, a system that is modifying driver behavior and nonrecurring congestion. capable of detecting weather and road surface conditions 4. Increased coordination with weather and traffic volume (e.g., rainfall amount, snow amount, visibility, wet road surface) data is required to determine when nonrecurring conges- and proposing possible road friction parameter variations tion exists and which driver actions result from these non- because of these conditions in order to issue corresponding recurring events. warnings, and a customized warning system initiated by the 5. It is possible to analyze naturalistic driving data to char- user's individual car key, which can adjust warning-issuing acterize typical levels of variability in travel times and threshold values according to different driving habits. develop measures for quantifying travel time reliability. When making countermeasure recommendations, it should be recognized that emerging driver assistance systems may Although the team has successfully proved the feasibility of initiate some complexities and, therefore, the assessment of using video and other in-vehicle and external data to study safety benefits is not straightforward. For example, when driver behavior and related nonrecurring congestion, some drivers rely on these safety systems, failure of such systems limitations need to be enhanced when a full data analysis is can be fatal. Incorporating some other countermeasures in conducted. These limitations are summarized as follows: a systematic approach will more than likely be beneficial. In conclusion, collision prevention should include a better 1. A limited number of safety-related events exist in the data design of roads, a more comprehensive recovery system, and sets the team examined because of the naturalistic nature a more coordinated safety management system. According to of the data. This shortcoming can be improved by extend- a report from the Organisation for Economic Co-operation ing the time duration of data collection or increasing the and Development, some basic enforcement may be highly number of participants. Both can be realized in the SHRP 2 efficient. Seat belt usage, speed management, extra efforts to Safety Project S07 study, in which a much larger data col- monitor high-risk drivers, and identification and monitoring lection effort will be performed. of dangerous locations are all effective countermeasures that 2. The external data sources in this study were examined only contribute to improvements in transportation system safety (8). for availability and accuracy. Because of time constraints, no real connection was conducted to relate driver behavior to external driving environment. Conclusions 3. Because of the limited size of travel time data in the natu- To determine the feasibility of using in-vehicle video data ralistic data sets, other data sets were also used to develop to make inferences about driver behavior that would allow travel time reliability models. These models are general investigation of the relationship between observable driver and apply regardless of the source of travel time data. behavior and nonrecurring congestion to improve travel time reliability, the team explored the identified data sets to inves- These limitations can be corrected if a larger video data set tigate the usefulness of video and other supplementary data, can be collected or the external data can be better linked with proposed models for the estimation of travel time reliability the in-vehicle data, which will be feasible in the next stage of measures, and identified potential problems in current data the study. sources. Based on the analysis of the six naturalistic data sources, this study demonstrates the following: Recommendations and Discussion 1. It is feasible to identify driver behavior before near crashes and crashes from video data collected in a naturalistic An important component of the next stage of SHRP 2 research driving study and thus infer the causes of those events. is a large-scale naturalistic driving data collection project 2. Recommendations can be made to change driver behavior (Project S07). The field study data collection contractors will and, therefore, prevent (or reduce) crashes and near crashes. be responsible for advertising for participants with preprepared 3. Naturalistic data are useful to identify impacts of crashes recruitment materials, scheduling participant drivers for on traffic conditions. Given the small sample of crashes installation and assessment, conducting driver intake testing and the fact that the DAS does not gather data when the and installing the DAS in the owner's vehicle, collecting data, engine is off, it is not possible to study the impact of inci- addressing problems encountered during the study, investi- dents on travel time reliability using in-vehicle data alone. gating crashes, transmitting data, carrying out quality control When effectively integrated with external data sources, procedures, and preparing periodic reports that document however, which is extremely feasible, given an accurate field study activities. The combined goal is to collect approx-
OCR for page 75
75 imately 4,000 vehicle-years of data in a 30-month period. The 27. Illuminance, Ambient: Ambient exterior light. following are the planned variables to be collected: 28. LDWS: Status of original equipment manufacturer (OEM) lane departure warning system. 1. Antilock Brake System (ABS) Activation: Antilock brake 29. Lane Marking, Distance, Left: Distance from vehicle activation indicator. centerline to inside of left side lane marker based on 2. Acceleration, x axis: Vehicle acceleration in the longitudi- vehicle-based machine vision. nal direction versus time. 30. Lane Marking, Distance, Right: Distance from vehicle 3. Acceleration, x axis fast: Vehicle acceleration in the lon- centerline to inside of right side lane marker based on gitudinal direction versus time. Fast buffer (-9 s to +3 s) vehicle-based machine vision. based on trigger (e.g., in crash or other high-acceleration 31. Lane Marking, Probability, Right: Probability that vehicle- event). based machine vision lane marking evaluation is providing 4. Acceleration, y axis: Vehicle acceleration in the lateral correct data for the right side lane markings. direction versus time. 32. Lane Marking, Type, Left: Type of lane marking imme- 5. Acceleration, y axis fast: Vehicle acceleration in the lateral diately to the left of vehicle using vehicle-based machine direction versus time. Fast buffer (-9 s to +3 s) based on vision. trigger (e.g., in crash or other high-acceleration event). 33. Lane Marking, Type, Right: Type of lane marking imme- 6. Acceleration, z axis: Vehicle acceleration vertically (up or diately to the right of vehicle using vehicle-based machine down) versus time. vision. 7. Acceleration, z axis fast: Vehicle acceleration vertically 34. Lane Marking, Probability, Left: Probability that vehicle- (up or down) versus time. Fast buffer (-9 s to +3 s) based based machine vision lane marking evaluation is providing on trigger (e.g., in crash or other high-acceleration event); correct data for the left side lane markings. 8. Airbag, Driver: Indicates deployment of the driver's 35. Lane Position Offset: Distance to the left or right of the airbag. center of the lane based on machine vision. 9. Alcohol: Presence of alcohol in the vehicle cabin. 36. Lane Width: Distance between the inside edge of the 10. Altitude, GPS: Altitude. innermost lane marking and the left and right of the 11. Audio: Audio recording for 30 s when incident button is vehicle. pushed. 37. Latitude: Vehicle position latitude. 12. Average Fuel Economy after Fueling: Average fuel econ- 38. Longitude: Vehicle position longitude. omy after fueling. 39. Pedal, Accelerator Position: Position of the accelerator 13. Cruise Control: Status of cruise control. pedal collected from the vehicle network and normalized 14. Date: UTC year, month, and day. using manufacturer specifications. 15. Distance: Distance of vehicle travel. 40. Pedal, Brake: On or off press of brake pedal. 16. Driver Button Flag: Flag indicating that the driver has 41. Pitch Rate, y axis: Vehicle angular velocity around the pressed the incident button. lateral axis. 17. Electronic Stability Control (ESC): ESC activation 42. Pitch Rate, y axis fast: Vehicle angular velocity around indicator. the lateral axis. Fast buffer (-9 s to +3 s) based on trigger 18. Engine RPM: Instantaneous engine speed. (e.g., in crash or other high-acceleration event). 19. Face, Driver ID: Machine-visionbased identification of 43. P-R-N-D-L: Gear position. the driver within those observed in a specific vehicle. The 44. Radar, Azimuth Forward: Angular measure to target. system observes within a vehicle to identify drivers who 45. Radar, Range Rate Forward: Range rate to forward radar drive that vehicle (i.e., not a unique identification across targets. all drivers in the study). 46. Radar, Range, Forward: Range to forward radar targets 20. Face, Gaze Zone: Estimation of the location of the driver's measured from the radar to the targets. gaze categorized into zones in and around the vehicle. 47. Radar, Target Identification: Numerical value used to 21. Face, Gaze Zone Confidence: Confidence in the estimation differentiate one radar target from others. of the zone at which the driver is looking. 48. Radius of Curvature, Machine Vision: Estimation of road- 22. Fuel Economy, Instantaneous: Instantaneous fuel way curvature based on machine vision. economy. 49. Roll Rate, x axis: Vehicle angular velocity around the 23. Fuel Level: Fuel level. longitudinal axis. 24. Heading, GPS: Compass heading of vehicle from GPS. 50. Roll Rate, x axis fast: Vehicle angular velocity around the 25. Headlight Setting: State of headlamps. longitudinal axis. Fast buffer (-9 s to +3 s) based on trigger 26. Horn Status: Actuation of horn. (e.g., in crash or other high-acceleration event).
OCR for page 76
76 51. Satellites, Number of: Count of the number of satellites mation by selecting certain check boxes. The before-trip being used for GPS position fix. information-collecting interface may consist of a list of the 52. Seat belt, Driver: Use of the seat belt by the driver. first names of household members for the driver to select 53. Speed, GPS: Vehicle speed from GPS. from, a list of trip purposes, weather conditions when the 54. Speed, Vehicle Network: Vehicle speed indicated on trip started, and any information about why the driver speedometer collected from network. selected the time of departure. The after-trip information- 55. Steering Wheel Position: Angular position and direction collecting interface may include an "original trip purpose of the steering wheel from neutral position. changed" option, a "route choice changed" option, and a 56. Sync: Integer used to identify one time sample of data "crash happened en route" option. Necessary hardware can when presenting rectangular data. be designed to connect the input touch-screen with the 57. Temperature, Interior: Vehicle interior temperature. engine so that the driver can start the engine only after 58. Time: UTC Time. Local time offsets need to be applied. the information is input. To ensure safety while driving, the 59. Track Type: Classification of target based on radar. device should be disabled while the vehicle is in motion to 60. Traction Control: Status of traction control system. prevent driver distraction. One major concern this type of 61. Turn Signal: State of illumination of turn signals. device may impose on such studies is that it will remind 62. Vehicle Angle Relative to Roadway: Vehicle angle relative drivers that they are being monitored and thus may reduce to the roadway based on machine vision. the naturalistic nature of the studies. 63. Video Frame: Frame number of video at point in time. 64. Video, Driver and Left Side View: Video capture of the Second, to serve the research purpose, certain data are more driver and exterior area to the left of the vehicle. important than others. The following four categories are 65. Video, Forward Roadway: Video capture of forward imperative: roadway. 66. Video, Occupancy Snapshot: Occupancy snapshot. 1. Basic onboard equipment should include devices that col- 67. Video, Rear View: Video capture to the rear of the vehicle. lect the following data: video; vehicle network information 68. Video, Right Side View: Video capture to the right of the (speed, brake pedal, throttle, and turn signal); GPS data vehicle. (latitude, longitude, and heading); X, Y, and Z acceleration; 69. Wiper Setting: Indicates setting of windshield wipers. distances between the subject and surrounding objects; 70. Yaw Rate, z axis: Vehicle angular velocity around the ver- lane location information (X, Y, and Z); driver behavior tical axis. (seat belt usage, lights on or off); and yaw rate. 71. Yaw Rate, z axis fast: Vehicle angular velocity around the 2. Video cameras should shoot at least five views: front, back, vertical axis. Fast buffer (-9 s to +3 s) based on trigger right, left, and the driver. The resolution of the video camera (e.g., in crash or other high-acceleration event). should be high enough to identify ongoing traffic conditions, weather conditions, and the driver's hand movements and To ensure that data collected in the SHRP 2 Safety Proj- facial expressions. Correction of sun glare to improve video ect S07 study are versatile and comprehensive enough to be quality is available when needed. used to conduct full-scaled research to study nonrecurring 3. The frequency setting should be high enough so that the congestion related to driver behavior, several recommendations video is continuous, the acceleration or deceleration of the have resulted from the findings of this study. vehicles should be clearly recorded, and the reaction times First, the procedure to recruit participants needs to be need to be recorded and measured. The recommended carefully designed. Ideally, a comprehensive population of minimum frequency for GPS devices is 1 Hz and, for all drivers ranging evenly across every age category, income other equipment, 10 Hz. category, and occupation category should be included. When 4. To improve the versatility of the data so that it can be recruiting participants, it is crucial to make it clear to them used in other, related research, the vehicle performance that driver information is vital for the research. To better parameters (e.g., engine speed, throttle position, and torque) identify drivers, two methods can be used: should be recorded. Table 8.11 shows a sublist of variables that are vital to the next stage of this research and that will 1. A formal statement needs to be included in the contract to be collected in the Project S07 study. Units and minimum make the signer the exclusive driver of the vehicle. rates of data collection are suggested. 2. A touch-screen device can be installed on board to collect information before and after each trip. The touch-screen Third, the data collection system needs to run for an addi- equipment can be designed so that a customized interface tional 10 min after the engine is turned off in case the vehicle will be displayed to the driver to input trip-related infor- is involved in an accident. During the additional data reduction,
OCR for page 77
77 Table 8.11. Recommended Variables for Collection data collection was usually found to stop the instant the driver stopped the vehicle. It is important, however, to observe the Recommended traffic conditions being affected by a safety-related event. In Variable Name Units Minimum Rate discussion with the SHRP 2 S06 contractor, a potential safety 1. Acceleration, x axis g 10 Hz hazard was identified that may deem this recommendation 2. Acceleration, y axis g 10 Hz infeasible. Specifically, continued data collection after an accident may result in a vehicle explosion if the vehicle gasoline 3. Acceleration, z axis g 10 Hz tank is jeopardized. 4. Altitude, GPS ft 1 Hz Fourth, to improve the linking of vehicle data with exter- 5. Date NA NA nal data, it is ideal to standardize the time and location data. 6. Distance mi NA For external data, the database in some states is built on the 7. Engine RPM rpm NA milepost system. The conversion of milepost locations to 8. Face, Driver ID NA 10 Hz standard latitude and longitude coordinates should be con- ducted ahead of time. For vehicle data, the synchronized GPS 9. Face, Gaze Zone NA 10 Hz clock should be used instead of the local computer time for 10. Fuel Economy, Instantaneous mpg NA better connection of the data with external traffic, crash, work 11. Heading, GPS degree 1 Hz zone, and weather data. 12. LDWS NA NA Fifth, because a limited number of crashes occurred in all 13. Lane Marking, Distance, Left ft 10 Hz the candidate data sets--especially severe crashes that affected 14. Lane Marking, Distance, Right ft 10 Hz traffic conditions--certain adjustments are needed to create a statistically significant database. A lengthier data collection 15. Lane Marking, Type, Left NA 10 Hz effort or more drivers involved in the study would be ideal. 16. Lane Marking, Type, Right NA 10 Hz For example, the 2,500-Car Study (SHRP 2 Safety Project S07), 17. Lane Position Offset ft NA which will soon be conducted, is a quality candidate. Another 18. Lane Width ft NA solution is simulation, which can be used to compensate for 19. Latitude Ddd.sss 1 Hz data shortage. Sixth, additional analysis of existing data is required to 20. Longitude Ddd.sss 1 Hz study typical levels of variability in driver departure times, 21. Pedal, Accelerator Position NA NA typical levels of variability in trip travel times, and the level of 22. Pedal, Brake NA NA variability in driver route choices. A characterization of this 23. Pitch Rate, y axis degree/s 10 Hz behavior is critical in attempting to quantify and develop travel 24. Radar, Azimuth Forward rad 10 Hz time reliability measures because it identifies potential causes 25. Radar, Range, Forward ft 10 Hz for travel time variability and thus can enhance travel time reliability models. These data may be augmented with tests on 26. Radar, Target Identification NA 10 Hz a driving simulator to study the impact of travel time reliability 27. Radius of Curvature, NA NA Machine Vision on driver route choice behavior. Finally, although a number of studies have used video cam- 28. Roll Rate, x axis degree/s 10 Hz eras to gather data, an ideal starting point is a compiled data 29. Seat belt, Driver NA 10 Hz source list that summarizes existing video-involved studies 30. Speed, GPS mph 1 Hz with specifications of data collected, limitations of data usage, 31. Time NA NA and access issues. Such a list would help prevent redundancy 32. Track Type NA NA in future investigation efforts. This research can benefit from the data being collected 33. Video Frame NA NA under the IntelliDrive Program (IntelliDrive is a service mark 34. Video, Driver and Left Side View NA 10 Hz of the U.S. Department of Transportation). The IntelliDrive 35. Video, Forward Roadway NA 10 Hz Program is, as introduced on its website, "a multimodal 36. Video, Occupancy Snapshot NA NA initiative that aims to enable safe, interoperable networked 37. Video, Rear View NA 10 Hz wireless communications among vehicles, the infrastructure, 38. Video, Right Side View NA NA and passenger's personal communications devices" (9). It will 39. Wiper Setting NA 1 Hz collect and disseminate data, including roadway, traffic condi- tion, weather, crashes, and traffic control among vehicles. With 40. Yaw Rate, z axis degree/s 10 Hz the development of IntelliDrive, it is possible to use the data