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.
3C h a p t e r 1 Naturalistic studies involve direct observation of events as they occur in natural settings. Naturalistic driving studies capture driver behavior in a way that is most representative of typical driving and not influenced by the artificial fea- tures of controlled studies. This method allows researchers to study drivers in their own vehicles and environments. Such data are expected to provide insight into factors that can influence safe driving. Recent advances in data collec- tion techniques have allowed capturing data on day-to-day driving behavior to be more feasible and cost-effective. Nat- uralistic driving behavior can be recorded with in-vehicle video cameras, and sensor arrays can detect brake and gas pedal changes, steering movements, and other vehicle and roadway factors. A central challenge of using naturalistic driving data- bases to identify the factors influencing driving safety is the selection of data analysis techniques that can address rel- evant research questions. In contrast to simulator studies, there is no experimental control in naturalistic studies. The causal mechanisms associated with safe driving are difficult to identify. Researchers must recognize and manage poten- tial confounders and covariates to reveal the influence of events or features of interest on driving safety. In the natu- ralistic setting, no two events are the same. For example, no two curves are the same. As a consequence, researchers must define equivalence classes to specify how a group of events can be considered as the same in an analysis. Defining equiv- alence classes requires that researchers define the driving context (e.g., road type, weather, traffic conditions, and time of day), as well as the specific parameter range (or threshold) that defines the event (e.g., equivalence classes for 500-ft- radius curves). To some degree, these equivalence classes substitute for the scenarios and experimental conditions that are used in driving simulator and other experimental studies. A key challenge in the analysis of naturalistic driving data is thus how to enhance the precision with which the researcher can infer the influence of driver, vehicle, and roadway characteristics on driving safety. Although naturalistic driving data can be used to analyze transportation safety and driving behavior in ways that were not previously possible, the spatial, dynamic, and temporal nature of the data adds to the complexity of such analyses. Sifting through the large volume of data collected can be extremely labor intensive and computationally difficult. Thus attention to methods of sampling, integration, and analysis is critical to reaching useful conclusions. The main goal of this project was to address the issues raised above in prepara- tion for Project S08âs analysis of the full-scale NDS. Project S02 had two primary objectives. The first was to identify and prioritize critical research issues related to driver safety (Phase I); the second was to determine the key research elements (e.g., methods, data, and questions) that will need to be addressed in analytical plans developed to explore these critical issues (Phase II). This report identifies a proposed set of high-priority research issues and presents a framework for developing work plans that include considerations related to data sampling and analysis. Five example work plans are provided. Introduction