Design of an intersection decision support (IDS) interface to reduce crashes at rural stop-controlled intersections

Abstract

Rural, stop-controlled intersections pose a crash risk to drivers, particularly elderly drivers. This paper outlines the design phase of an infrastructure-based intersection decision support (IDS) system to help drivers make safer gap acceptance decisions at rural intersections. A human factors-based design process was conducted to determine the type of information that should be presented to drivers. Information considered important for presentation to the driver included showing the presence of gaps, indicating the size of available gaps, and/or judging the safety of available gaps. This paper discusses the process used to determine the appropriate design specifications for initial testing of the IDS system interface.

Introduction

Some states in the US are currently investigating the problem of rural intersection negotiation in an effort to develop intelligent transportation system (ITS) countermeasures to reduce fatal crashes on rural roads. In particular, rural stop-controlled intersections where high-speed, high-volume roads (termed Interregional Corridors or IRCs) are intersected by lower-speed, lower-volume roads controlled by a stop sign are a major problem (see Fig. 1). In 2002, over 22,000 fatal crashes occurred on rural roads in the US, accounting for 59% of all fatalities (NHTSA, 2003). According to the American Association of State Highway and Transportation Officials (AASHTO) (1997) approximately 16% of rural fatal crashes in the US occur at intersections. At the state level, there were 34,175 reported crashes on Minnesota rural two-lane roads between 2000 and 2002 (Preston and Storm, 2003). Over 32% (11,069) of these Minnesota crashes were intersection related, with 22% of the fatal rural accidents occurring at stop-controlled intersections.

The issue of limited or obscured sight distances has been addressed through the use of actuated warning systems in previous research (e.g., Hanscom, 2001, Peabody et al., 2001). However, there a number of factors, including poor sight distances, that affect rural stop-controlled intersection negotiation. Therefore, a human factors effort was undertaken in Minnesota to specify an appropriate design for an infrastructure-based intersection decision support (IDS) system to aid drivers at stop-controlled rural intersections. An infrastructure-based solution provides a practical starting place for the development of an IDS system because the system can be viewed by all drivers when they arrive at the intersection. Depending on its final design, an infrastructure-based system could be modified in the future to cooperate with in-vehicle systems as those become more prevalent in vehicles. A non-cooperative system must assume general parameters to be applied to all drivers, whereas the advantage of a cooperative system is to communicate information about the vehicle and driver (e.g., age) to the infrastructure to provide more specific information.

A task analysis, which is a commonly used tool to identify driver behavior in specific situations (e.g., Caird and Hancock, 2002), was conducted in conjunction with a review of crash data and driver errors at rural stop-controlled intersections to identify contributing crash factors. These factors were used to constrain the design space and highlight information that may be critical for drivers to make safe decisions at this type of intersection. The overarching goals of the system are to focus on the minor-road driver’s behavior and how best to support their decision making to enter the intersection safely and complete their intended maneuvers (i.e., crossing, turning). Once the task and error analyses were complete, design criteria were developed and prototype IDS systems were proposed then evaluated by subject matter experts (SME) in the transportation field.

A task analysis provides the framework necessary for synthesizing and interpreting the results of crash data. Previous intersection task analyses examined single aspects of intersection negotiation (e.g., left turns) (Chovan et al., 1994a, Chovan et al., 1994b, Tijerina et al., 1994) or applied the analysis broadly to all intersection types (e.g., Caird and Hancock, 2002). Because the proposed IDS system will be for use specifically at rural, stop-controlled intersections it was necessary to tailor the analysis specifically to the tasks required to navigate that type of intersection. The rural stop-controlled intersection task analysis identifies task goals, tasks, and sub-tasks that encompass the range of perceptual, cognitive and behavioral aspects of negotiating the intersection. Task goals group similar tasks together, while tasks are further broken down into sub-tasks that are necessary for accurate completion of the task. Table 1 shows the task goals, tasks and sub-tasks identified for rural stop-controlled intersection negotiation.

An analysis of rural stop-controlled intersection crashes helped identify which tasks were providing drivers with the most problems. Previous research has identified gap acceptance problems as a significant contributor to rural intersection crashes (Chovan et al., 1994b, Najm et al., 2001). Additional crash information from Minnesota supports the previous findings that gap acceptance is a problem at stop-controlled intersections. For example, in Minnesota, right-angle crashes associated with turning maneuvers from the minor roadway accounted for 18.7% of two-lane roadway crashes and 21.1% of crashes on IRCs (Preston et al., 2004). Detailed analyses of crash reports at stop-controlled intersections in Minnesota also showed that 57% of drivers on the minor road stopped before proceeding and were struck in the intersection by oncoming traffic (Preston and Storm, 2003). At one problem intersection where 65% of accidents were right-angle collisions involving a major-road vehicle and a minor-road vehicle, approximately 87% of the drivers involved in these crashes stopped before entering the intersection.

Previous research and the current Minnesota crash analysis imply that the majority of drivers detected the intersection (Task A) and stop sign (Task E) appropriately but had problems entering the intersection safely. Failures to detect approaching vehicles that form a gap (Task H), to perceive or estimate the size of the gap (Task I), or to appropriately judge the gap as safe (Task J) may all result in serious, right-angle collisions. Therefore, the IDS system will support the gap acceptance tasks as these are critical for safely negotiating the intersection.

Although data indicate that some drivers also failed to stop at the intersection, the goal of the IDS program is to support decision making for drivers who stop as this has been identified as the main problem in Minnesota and in previous research (Chovan et al., 1994b, Najm et al., 2001). Further, the limitations of a non-cooperative, infrastructure-based system constrain the ability to address the issue of deliberate violators or distracted drivers who miss the intersection. Drivers who habitually violate stop signs are not likely to respond to the infrastructure solution if they already do not respond to legal traffic signals present at the intersection. Moreover, the goal of the IDS system is to support driver decision making. Violators who decide to run the stop sign are not necessarily interested in making a safe crossing decision using decision support technologies. Drivers who miss detecting the intersection or stop sign (Tasks A & E) due to distraction may benefit from the presence of the additional signage associated with the IDS system at the intersection. That is, the presence of additional infrastructure may draw a distracted driver’s attention to the intersection. Future iterations of an IDS system may communicate cooperatively with in-vehicle systems to warn distracted drivers (or potential violators) of the intersection’s presence.

Previous research has identified a number of errors associated with the gap acceptance tasks in Table 1 (Tasks H, I, J) that precipitate crashes at intersections (Caird and Hancock, 2002, Chovan et al., 1994a, Chovan et al., 1994b, Staplin et al., 1998). Table 2 highlights these errors and how they apply to the tasks of rural stop-controlled intersection negotiation. Tasks that cannot be addressed by a non-cooperative infrastructure solution but also lead to serious collisions, such as failure to stop or decelerate properly, are not included in Table 2.

Research shows that problems associated with detecting vehicles approaching on the major road are common at stop-controlled intersections. Drivers often report “looking but not seeing” a vehicle after a crash, which is most likely due to inattention (Treat et al., 1979, Chovan et al., 1994a, Chovan et al., 1994b) or perceptual problems associated with age or disease (Staplin, 1995). Limited sight distances at the intersection may also impair a driver’s ability to see approaching vehicles (Preston and Storm, 2003). Therefore, the IDS system should draw attention to approaching vehicles to better assist the driver with the gap acceptance task.

If a driver is able to detect the approaching vehicles and their associated gaps, they must then form a perception of gap size (Task I). To assess gap size, drivers must have a sense of how fast the approaching vehicle is traveling, how far away the vehicle is, and how soon it may arrive at the intersection. Scialfa et al. (1987) showed that drivers, on average, estimated that vehicles were traveling slower than they actually were. In particular, drivers were more likely to underestimate how fast a vehicle was traveling as the speed of the approaching vehicle increased. Hancock and Manser (1997) observed that older participants showed greater underestimation of approach speed overall and were less accurate than younger drivers as approach speed increased. Underestimation of approach speed could result in overestimation of arrival time as vehicles are perceived to be traveling slower than they actually are. Hancock and Manser (1997) also found that when actual arrival time was larger and approach speeds increased, all participants overestimated arrival time. These results suggest that overestimation of arrival time may be more likely when approach speed increases, actual arrival times are larger, and vehicles are further away. Overestimation of arrival time, which may be due to an underestimation of approach speed, could lead to a perception of gap size that is larger than reality and drivers may accept gaps that are less safe. This interpretation is consistent with the rural intersection crash data and suggests drivers have problems perceiving gaps for vehicles that are far away and traveling at high velocities (such as on rural highways).

Once a driver has formed a perception of gap size, they must decide whether the size of the gap is sufficient to safely enter the intersection and complete their intended maneuver. In addition to the vehicle’s approach speed and arrival time, several other factors may influence a driver’s decision to accept a gap. These factors include:

• the type of vehicle the driver is driving (small versus large);

• driver age;

• the intended maneuver (left turn, right turn, straight crossing);

• the number of lanes to cross;

• the road surface conditions.

Fitzpatrick (1991) showed that larger gaps are accepted by drivers of larger vehicles to compensate for reduced accelerative capabilities and longer vehicle lengths. Lerner et al. (1995) found that older drivers accepted larger gaps relative to younger drivers. Regardless of other factors, drivers making left turns were shown to consistently accept larger gaps and that drivers accept larger gaps when there are more lanes to cross (Harwood et al., 1999). Finally, Cooper and Zheng (2002) found that larger gaps were accepted when the pavement was wet compared to when it was dry. All these factors should be considered when designing the gap threshold to be used in an IDS system.

When the above factors that influence gap detection, perception and acceptance are taken into consideration, older drivers present the highest-risk group of drivers for these tasks. Older drivers are over-represented in rural intersection collisions (Staplin and Lyles, 1991, Stamatiadis et al., 1991, Preusser et al., 1998). In Minnesota, a safety audit of Highway 52 (location of the IDS test intersection) found that older drivers accounted for 39% of the rural intersection crashes compared to a state-wide average of 14% (Preston and Rasmussen, 2002).

Older drivers have problems with perception, cognition and attention that can affect their ability to adequately detect, perceive and accurately judge the safety of a gap. Problems with divided attention and attending to only one source of information (Dewar, 2002, Hakamies-Blomqvist, 1996, McDowd and Shaw, 2000), visual search (Ho et al., 2001, Maltz and Shinar, 1999), and narrowing of the useful field of view (UFOV) (Ball and Owsley, 1991) are common among older adults, and could contribute to errors in detecting approaching vehicles (Task H). Research also shows older drivers are especially prone to the underestimation of approach speed (Hancock and Manser, 1997, Scialfa et al., 1987), a skill necessary for accurately perceiving and accurately judging the safety of a gap (Tasks I and J). Although older drivers frequently accept larger-sized gaps than younger drivers (Lerner et al., 1995), they also accelerate more slowly (Hakamies-Blomqvist, 1996), take more time to make turn decisions (Caird et al., 2002), and require more time to complete maneuvers (Parsonson et al., 1996) (Task K). As is the case for other drivers, left turns are particularly difficult for older drivers and typically require more time to complete (Caird and Hancock, 2002). Therefore, although older drivers accept larger gaps, these gaps may not be large enough to completely eliminate accident risk at intersections (Alexander et al., 2002).

Based on the analysis, a number of design guidelines and constraints were identified for a rural stop-controlled IDS system. The first constraint on the system is the need to assume safety thresholds based on an older driver making a left turn. This is because the IDS system considered here is non-cooperative, thus it will not know what maneuver the driver intends or the age of the driver at the intersection. Older drivers require the largest gaps, particularly for left turns, and this must be accounted for by the system.

A second constraint on the system is to support the minor-road driver’s behavior with minimal interference on the major-road driver’s behavior. According to AASHTO (2001) guidelines for intersection sight distances, safe gaps are based on the assumption that major-road vehicles decelerate up to 30% to avoid a collision. The IDS solutions would assume the same limits by intending not to reduce the speed of major-road traffic by more than 30%.

Third, each proposed interface is a technically feasible solution, which could be applied to all rural, stop-controlled intersections. Experimental technology is not considered. When possible, the solutions assume or estimate parameters, such as driver age and perception-response time, to avoid adding additional technology that would measure them directly. To further minimize complexity, gaps will be calculated from the minor-road driver to the next closest vehicle on the major road. This metric is of greatest interest to the driver because they are most concerned with the time or distance to the nearest approaching vehicle, not those that have already passed that make up the gap. A sensor system is currently deployed at a candidate intersection in Minnesota that can be adapted for future use with an IDS system (Alexander et al., 2005).

Finally, an IDS system can be designed to support different types of turning strategies. A one-stage strategy is when a driver accepts gaps in order to complete the left-turn maneuver without stopping in the median. In this case, the driver must ensure that both the near and far gaps are large enough to cross the near lanes, the median and turn into the far lanes. A two-stage strategy occurs when drivers first accept the near-lane gap to reach the median, then stop in the median and evaluate the far-lane gaps independently for an appropriate gap in which to enter the far lanes of traffic. There is no research to quantify the prevalence of choice for either strategy or the relative risk of a one-stage versus a two-strategy. A design goal was to support both types of strategies where possible.