|
Article Excerpt
INTRODUCTION
In 2001 there were about 216,000 public road backing crashes, resulting in 18,000 injuries and 140 fatalities (National Highway Traffic Safety Administration [NHTSA], 2002). The actual number of backing collisions is difficult to estimate because these types of accidents tend to occur in areas (parking lots, driveways, private roads, etc.) explicitly excluded from NHTSA's General Estimates System (GES) and Fatality Analysis Reporting System (FARS); both the GES and FARS databases address crashes on public roads and not parking lots or private roads and driveways. Some estimates suggest that the actual number of backing-related crashes could be much larger--about 500,000 annually, resulting in 50,000 injuries and 390 fatalities (NHTSA, 1997; Wang & Knipling, 1994).
Children are represented in higher numbers than adults in these types of accidents. In July 2000 through June 2001, an estimated 2767 children 14 years old and younger were run over/ backed over by a motor vehicle in a nontraffic-related incident. An estimated 27% of nontraffic-related fatal injuries of children in or around motor vehicles are a result of being backed over (Centers for Disease Control, 2002). Children 1 to 4 years of age are particularly vulnerable because their height makes them more difficult to see or because they do not understand the danger. This age group represents 6% of the U.S. population, yet it accounts for 30% of all off-road backing fatalities (NHTSA, 1997).
Several automobile manufacturers offer backing/parking aids designed to assist drivers while backing. These systems may include side-view mirrors that rotate downward when the vehicle is placed into reverse as well as Ultrasonic Rear Park Assist systems to help locate fixed objects behind the vehicle near the bumper. Proximity-based parking aid systems are intended to help the driver determine how close an object is to the bumper within a limited operating range (e.g., within 1.5 m of the vehicle) during low-speed (e.g., <5 km/h) parking/backing conditions. Unlike parking aids, back-up warning systems are intended to warn drivers of the presence of unexpected or unseen objects behind their vehicles at relatively higher backing speeds.
To be effective, a back-up warning system must capture the attention of an unalerted driver and provide sufficient advance notice to enable drivers to respond appropriately. The warning time needed includes system detection and processing time, driver perception-reaction time, and vehicle maneuvering (stopping) time. In approximate terms, system sensing and processing times are often very short (hundreds of milliseconds) once the object has entered the system's field of view. Driver perception-reaction time typically varies in the 1- to 2-s range, depending on a number of variables, including saliency of the system interface and individual differences among drivers (Green, 2000). Backing speeds are widely distributed depending on the situation and maneuver; during parking, speeds can range from 3 to 7 mph (4.8-11.3 km/hr; Huey, Harpster, & Lerner, 1995; Kiefer, Llaneras, & Lerner, 2001). Braking levels in response to a backing alert have been observed to vary from 0.1 to 0.5 g, averaged over the duration of the event (Kiefer et al., 2001).
Thus back-up warning systems need to provide advance warning to the driver in order to be effective and must also cover different backing situations that occur at a range of speeds. Furthermore, back-up warning systems must also distinguish between situations in which the driver intentionally backs close to known objects and situations in which the driver is unaware of an in-path obstacle. The former situation complicates the implementation of an effective warning system by restricting the potential for nuisance warnings; too many nuisance alerts may cause drivers to ignore the warning or delay responding to real threats.
In order for a rear obstacle detection system to help the driver avoid a collision with a person (or object) during a backing maneuver, the driver must be provided with reliable information in a salient manner in sufficient time to take preventive action. Current ultrasonic parking assist systems provide a limited detection range and are prone to a high number of false alerts; this is why these features are appropriately used only for parking tasks. Radar-based technology and ultrasonic-radar hybrids may provide sufficient range, but commercial sensor systems are only now becoming available. Current and proposed systems constrain the sensors' field of view both laterally and longitudinally in order to mitigate false alerts, and the detection zone may contain blind spots near the vehicle because of the physical characteristics of the vehicle's sensors. As a result, systems are limited in their ability to detect objects and to provide drivers with sufficient warning time under all possible backing scenarios. Even if capable sensing technology were available, the issue of driver behavior in response to an alert from a rear obstacle detection system remains.
Guidelines for the design of backing warning systems recommend the use of multiple warning levels that include cautionary and imminent crash alerts (Huey et al., 1995; Lerner, Kotwal, Lyons, & Gardner-Bonneau, 1996). Lerner et al., for example, recommended that a back-up warning device should include both a cautionary and a danger warning signal based on time to collision (TTC), with the former occurring at about 3.5 s before impact and the latter at about 1.5 to 2.0 s. The warning system should be primarily acoustic and could be supplemented by a visual display located over the driver's right shoulder, which has been found to be the dominant glance direction when backing (Huey et al., 1995). Huey et al. (1995) also indicated that back-up warning systems should consider operation within a variety of backing situations and take into account the different relationships between the vehicle and the potential collision object. This includes scenarios in which the vehicle is backing toward an object in its path; the object enters the path of the backing vehicle; and the object is in very close proximity to the rear of the vehicle prior to the initiation of backing.
Key to driver acceptance of backing warnings (and crash avoidance alerts in general) is appropriate crash alert timing, which refers to the necessary underlying kinematic (or approach) conditions for triggering the onset of crash alerts. As stated earlier, the goal of a crash alert timing approach is to allow the driver enough time to avoid the crash and yet avoid annoying the driver with alerts perceived to occur too early (i.e., under nonthreatening conditions). To develop appropriate timing of the crash alerts, one needs to consider two fundamental driver behavior parameters (Kiefer et al., 1999). For back-up warning systems, the first driver behavior parameter is driver brake reaction time (RT; i.e., the time between crash alert onset and the driver triggering the brake switch), and the second parameter is the driver deceleration (or braking) behavior in response to the alert (which determines stopping distance or stopping time). These parameters serve as input into kinematic equations that determine, given a set of assumptions, the range to the object at which an alert is necessary to assist the driver in avoiding a potential backing crash.
This second parameter has recently been examined by Kiefer et al. (2001). The 96 drivers in that backing study were asked to brake at the last possible second in order to avoid colliding with an obstacle under various speed and braking intensity instructions, as well as under various maneuver conditions. This experimental strategy was employed to develop a fundamental understanding of the timing and nature of drivers' "last-second" braking behavior without a back-up warning system, so that "hard braking" kinematic situations could be properly identified and modeled for crash alert timing purposes. Such an experimental strategy has been successfully used to develop a timing approach for a forward collision warning system (Kiefer, 2000; Kiefer et al., 1999, 2003).
This study assessed the appropriateness of warning timings based on a model developed from the Kiefer et al. (2001) study (coupled with brake RT assumptions), as well as production-friendly interface approaches of a prototype rear obstacle detection system (RODS) using a different sample of drivers. Specific characteristics evaluated included the warning timing algorithm and the system interface. More specifically, the visual and auditory warnings evaluated were fully integrated with the interface currently employed in the Ultrasonic Rear Park Assist system offered by General Motors on a wide range of production vehicles. Two warning algorithms and three basic interface approaches were initially tested; a fourth interface scheme was added to the design following data collection with one of the study vehicles. Driver response to the warning system...
|
