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.
5Research Questions For decades crashes have been studied as discrete events focusing on the circumstances of the crash event. This type of analysis is exemplified in the review of Appendix A, and has been used to identify the characteristics of highway features associated with higher crash experience; other factors such as traffic volumes, driver characteristics, land use, and the envi- ronmental conditions were also needed to explain or describe crash events. Furthermore, cross-sectional analyses of crash events did not address circumstances leading to a crash. Advances in vehicle instrumentation technology have made it possible to collect longitudinal naturalistic data about the vehicle, driver, and roadway, accumulating information about events preceding a crash, if a crash occurs. However, crashes are rare events, and there are conditions in which a crash, while likely, does not occur. Thus a crash can be considered as a high-probability outcome, given a set of conditions, and validated crash surrogates could be used to identify these con- ditions. Assuming that no driver intentionally crashes, it fol- lows that crashes occur when there is loss of situational control leading to a damaging impact, and hence that surrogates are related in some way to disturbances of the driving control function. Also, it should be possible to identify the disturbance of control from NDD. The research team also expects that crashes are related to crash surrogates in an objective way that the team seeks to determine. âControlâ is defined here as the effectiveness of tactical and operational aspects of the driving task (i.e., acquiring and tracking reference information for speed and steering adjust- ment). âDisturbed controlâ is any interruption or delay in the process of perception (seeing lane boundary or other relevant features that determine the required path), recognition (what are the relevant objects that are relevant to speed and steering control?), judgment/decision (of required steering, throttle pedal, or brake pedal) or action (apply corrections) in the driving task. Disturbed control is not expected to be the same as poor lane keeping. It is quite common in NDD to see lane excursions, such as âcutting a curveâ or use of the shoulder, in which the driver appears fully aware of the action and is simply not tracking within the lane markings. One might argue that these excursions still represent poor control (i.e., they do not conform to the transportation researcherâs expectations), but if the driver decides to manipulate the reference conditions used for steering control (essentially a tactical decision to, for example, cut the curve) and follows that action with accuracy and predictability, then at least at the operational level, the con- trol loop is effective. The risk of such behavior clearly depends on the skill and awareness of the driver. Because the number of factors associated with vehicle crashes increase significantly if more than one vehicle is involved, this research examines only the single-vehicle road departure crashes (i.e., crashes involving only one vehicle in which the first harmful event occurs off the road- way). Thus, the research team tentatively expects that crash surrogates are related in some way to the disturbance of the control function of the driving task, and that it is possible to identify various types of disturbance of control from naturalistic driving data. The research team also expects that crashes are related to crash surrogates. These general considerations are now formalized as research hypotheses as follows. Research Hypotheses The research hypotheses are as follows: 1. Single-vehicle road departure crashes occur only under conditions of disturbed control. 2. Naturalistic driving data contain measurable episodes of disturbed control. 3. Crash surrogates exist and are based on a combination of objective measures of disturbed control (from onboard sensors), highway geometric factors, and off-highway factors (environmental factors). 4. Crash surrogates can be related to actual crashes. C H a p t e R 2
63. Are other driving control metrics necessary (in addition to vehicle kinematic measures) to identify disturbed control? Relating Driving Performance to Geometric Features and Road Departure Crash Frequencies 4. Are there measures of driving control performance in existing field operational test (FOT) data that depend on highway factors in a way that is consistent with single- vehicle road departure crash frequencies? 5. Are there specific highway features that are associated with single-vehicle road departure crashes and specific driving control performance measures? (Possible candi- dates are isolated horizontal curves, sharp horizontal curves, sequences of horizontal curves, and combinations of horizontal and vertical curves.) 6. Can roadside factors (e.g., locations of poles, trees, bridge abutments, and side slopes) be coupled to naturalistic driving data? 7. Does the coupling of roadside factors to naturalistic driving data improve correlation with actual crashes? 8. Can general descriptors of roadside environments be used in this coupling (e.g., tree density and proportion of side slope steeper than 4 to 1), or do we have to be more specific about location of roadside obstacles? Statistics 9. What statistical tests are available to determine if the measures of driving control performance in naturalistic data and single-vehicle crashes depend on geometric fea- tures in a consistent way? 10. Can satisfactory crash risk predictions be made on the basis of vehicle/driver/highway information available from nat- uralistic driving (e.g., via extreme value theory), or do additional roadside and environmental factors need to be introduced? Driver Factors 11. Is the pattern of driving control performance different for the same driver when distracted versus not distracted (e.g., on a cell phone or not on a cell phone)? 12. Can various driver states (e.g., drowsy, aggressive, dis- tracted, engaged) be identified from naturalistic driving data? 13. Can driving control performance for various states be categorized more simply (i.e., good and bad, or risky and nonrisky)? 14. Is there a difference in the driving control performance of good and bad drivers (or risky and nonrisky drivers) at locations with geometric features associated with high single-vehicle crash frequency? When conditions leading to single-vehicle crashes are con- sidered, the research team expects that in many cases the dis- turbance to the control function might be measurable over an extended period of time; for example, a drowsy driving surro- gate would only emerge as significant over time as lane-control dysfunction was found to be persistent, compared to what might be seen in a short period of distraction. The key seems to be that surrogacy is an indicator of extremes in a uniform process that includes crashes at the limits, and therefore a road departure surrogate includes the crucial element that it mea- sures how the driving control loop is disturbed and is not sim- ply being manipulated by the driver. While this is a useful guiding principle for the definition of surrogates, it is a matter for analysis and verification of how well such surrogates are matched to actual crash data. The intention has been to focus on those questions most directly related to the hypotheses of disturbed control, surro- gacy, and relationships between data. With this focus, research questions can be posed at many levels from broad general ques- tions down to very specific direct technical questions. The research team focused on three levels: the first level was a restate- ment of the research hypotheses, the second level was specific questions of safety research, and the third level was data quality and validation. Broad Research Questions The research questions are summarized as follows: 1. Do single-vehicle road departure crashes occur only under conditions of disturbed control? 2. Do naturalistic driving data contain measurable episodes of disturbed control? 3. Do objective measures of disturbed control from naturalis- tic driving data, together with highway geometric factors, off-highway factors, and environmental factors, satisfy crite- ria for crash surrogate (i.e., are they related to actual crashes)? Specific Safety Research Questions Specific research questions are broken down into four subtypes: measuring disturbed control, relating driving performance to geometric features and road departure crash frequencies, statis- tics, and driver factors. Measuring Disturbed Control 1. What measures exist in naturalistic driving data that directly measure disturbed control? 2. Are vehicle kinematic measures sufficient to identify dis- turbed control for risk measures in single-vehicle road departure crashes?
7data, and what are the levels of accuracy in those measures? 2. What spatially referenced crash and highway data exist in the regions where the driving took place, and what gaps exist in the data? 3. Can the analysis of data in southeastern Michigan be applied or recreated in another region (e.g., Virginia)? Data Quality and Validation A number of lower-level research questions are related to crosschecking and data quality: 1. What kinematic measures of driving control perfor- mance are available in the available naturalistic driving
