U.S. Department of Transportation
Federal Highway Administration
1200 New Jersey Avenue, SE
Washington, DC 20590
Federal Highway Administration Research and Technology
Coordinating, Developing, and Delivering Highway Transportation Innovations
|This report is an archived publication and may contain dated technical, contact, and link information|
Publication Number: FHWA-HRT-06-033
Date: August 2006
Task Analysis of Intersection Driving Scenarios: Information Processing Bottlenecks
PDF Version (4.01 MB)
PDF files can be viewed with the Acrobat® Reader®
Scenario 3–Straight on Yellow Light
This scenario involves the subject vehicle going straight through the intersection on a yellow light. Figure 22 shows the scenario diagram and provides additional details regarding the scenario. Briefly described, this scenario involves the driver approaching the intersection just as the light turns yellow. The driver is assumed to be in the dilemma zone where it is somewhat ambiguous as to whether the safest course of action is to stop or to go through the intersection. In this scenario, the driver does decide to proceed.
This scenario is divided into three segments (Approach, Decision to Proceed, and Intersection Entry). Only three segments were required in this scenario because for most of this scenario, drivers are mostly performing similar tasks related to traveling forward, with the decision to proceed briefly punctuating this process in the middle. The primary basis for parsing this scenario into these segments is that each segment had a different overall driving goal (table 31). Unlike with other scenarios, the speed of the subject vehicle does not vary greatly between segments. Figure 22 shows the diagram for Scenario 3 and lists Scenario 3 activities.
The crash data regarding this scenario indicate that the decision to proceed likely represents a source of problems for many drivers.(10) In particular, the most common contributing factors include deliberately running the signal (40 percent), either because drivers failed to obey the signal (23.1 percent) or tried to beat the signal (16.2 percent). The next most common contributing factor was driver inattention (36.4 percent). There are two explanations of why these problems are linked to the decision stage.(10) In particular, the decision stage essentially involves assessing the situation and then deciding whether it is safe to go based on that assessment. Thus, there are two ways for drivers to end up running the signal when the appropriate action is to stop: correctly assessing the situation as unsafe and then making a bad decision to go anyway, or incorrectly assessing the situation as safe (because the driver missed relevant information) and making the logical-but incorrect-decision to proceed. The latter case is similar to driver inattention, whereby drivers also fail to adequately perceive and process the necessary situational information.
Several assumptions were made regarding the situational aspects of the scenario. The justifications for these are described in table 32 and summarized in figure 22.
An approximate timeline showing the key temporal milestones for Scenario 3 was calculated based on vehicle kinematics (figure 23). These milestones were used to make judgments about the pacing of tasks within segments and provide a basis for the overall sequencing of certain tasks. Most segments included an interval with a variable time component, which represented intervals that were either long enough to effectively provide unlimited time to perform tasks or of a duration that was determined external to vehicle kinematic factors (e.g., waiting for lead vehicle to turn).
of key segment phases and event/task milestones.
The results of the task analysis organized by scenario segment are shown in table 33, the task analysis table. The task analysis results are duplicated for individual segments in the segment analyses tables in the next sections, which more fully discuss the organization and content of the tasks and information processing subtasks.
Table 33. Scenario 3–Straight on Yellow Light task analysis table, continued.
The Approach segment involves the subject vehicle traveling at full speed until the traffic signal turns yellow. The tasks, information processing subtasks, and workload estimates associated with this segment are shown in table 34. The scenario diagram, relative timing of tasks, and potential contributions to information processing bottlenecks and mitigating factors are shown in figure 24 and table 35.
tasks and information processing subtasks.
Task 3.1.7 (determine if the light is about to change) can range in difficulty based on the degree to which direct information is or is not available. For example, if a pedestrian walk signal were visible during the approach, it could make the perceptual task into a simple inspection task (workload demand = 3) rather than a potentially complicated visual search task (workload demand = 7) to identify cues distributed throughout the roadway (e.g., looking for a vehicle queue in oncoming traffic or observing pedestrian behavior). The more demanding version was selected in keeping with the overall goals of addressing the worst-case scenarios. Note also that task 3.1.6 (observe status of the light) is intended to reflect early viewing of the signal that is part of the process of initially assessing the situation, and not the glance that observes the light changing to yellow.
Task Pacing and Timing - Task 3.1.1 (lane maintenance) is forced-paced because it is part of the ongoing task of driving, but the other tasks are self-paced.
Regarding the task ordering, tasks 3.1.4 through 3.1.7 are sequential and can essentially be performed in any order. The ordering chosen in this segment followed a logical sequence and was consistent with the ordering of similar tasks in the Approach segment in the other scenarios.
The Decision to Proceed segment spans from the time that the traffic signal turns yellow to when the driver decides to proceed, which occurs within a few seconds. The tasks, information processing subtasks, and workload estimates associated with this segment are shown in table 36. The scenario diagram, relative timing of tasks, and potential contributions to information processing bottlenecks and mitigating factors are shown in figure 25 and table 37.
tasks and information processing subtasks.
*Difficulty in this subtask is increased by a value of 1 because of degraded information.
There are several noteworthy points in this segment. The first is that the lane maintenance and hazard monitoring tasks (3.2.1 through 3.2.3) are included because even though many drivers are unlikely to perform these tasks in the relatively brief time it takes to make their decisions, they are nonetheless still required for safe driving and including these tasks emphasizes that drivers are more likely to skip these tasks. Also, the estimated workload value for task 3.2.6 (determine if stopping will conflict with following vehicle) cognitive subtask was incremented by a value of 1, because determining the trajectory of the following vehicle is more difficult to do using the degraded indirect visual information from the rearview mirror.
Finally, there is no separate task defined for the actual decision about whether to proceed. This decision process is the culmination of tasks 3.2.5 through 3.2.8, and the assumption is that the decision is an implicit result of resolving these tasks. The exception would be if the precursor tasks yielded contradictory decisions (e.g., the lead vehicle is stopping, but stopping will result in a conflict with the following vehicle). In this case, a separate decision task would be necessary to resolve the conflict. In the present scenario, however, the precursor tasks do not result in contradictory decisions, and these conflicts do not exist.
Task Pacing and Timing - Task 3.2.1 (lane maintenance) is forced-paced because it is part of the ongoing task of driving. Tasks 3.2.4 through 3.2.8 are forced-paced because they have to be completed in a very limited amount of time (approximately 1 sec in this scenario).
Regarding the task ordering, tasks 3.2.5 through 3.2.8 were depicted as occurring simultaneously when they must in fact be performed in a sequential manner. These tasks were shown this way because they must be performed in an exceptionally short time span, which may result in interference between the tasks. In addition, depicting these tasks in this way emphasizes the fact that it is unlikely that drivers are able to complete all of these tasks in the available time. Note also that the order of tasks 3.2.5 through 3.2.8 likely varies between drivers. Task 3.2.8 (determine if oncoming vehicles are trying to turn left) was placed at the end because drivers probably have more flexibility regarding whether to proceed based on this information than the other tasks. Although stopping for left-turn traffic is the appropriate action, failing to do so is less likely to result in a vehicle conflict than making an inappropriate decision in tasks 3.2.5 through 3.2.6.
One important point that the task analysis underscores in this scenario is that dilemma-zone situations provide limited options for drivers.(23) In the current example, if drivers spend more than 1 sec performing the tasks described in this segment, they only have the option of continuing through the intersection, because after 1 sec they are too close to stop comfortably before the stop line (even with aggressive values for response time and deceleration). Note that this also represents an optimal situation because the subject vehicle is assumed to be in the limited option zone region in which it is possible to either stop or legally enter the intersection. In contrast, if the vehicle is farther back and the driver takes a little longer to decide what to do, then the driver may be left with no safe or legal options. For example, if the subject vehicle is just 10 to 15 meters (m) farther back when the light turns yellow, and if the driver takes 2 sec rather than 1 sec to complete decisionmaking tasks, then the vehicle will be too close to the intersection to stop, yet too far from the intersection to enter before the light turns red without accelerating (arrival will occur more than 0.5 sec after the light turns red). Thus, this is an inherently difficult segment for drivers, not only because they have an extremely limited amount of time to perform several tasks, but they may also be limited in the types of actions they can safely or legally take.
The Intersection Entry segment spans the interval from when the driver decides to proceed into the intersection to when the subject vehicle crosses to the other side of the intersection. The tasks, information processing subtasks, and workload estimates associated with this segment are shown in table 38. The scenario diagram, relative timing of tasks, and potential contributions to information processing bottlenecks and mitigating factors are shown in figure 26 and table 39.
of tasks and information processing subtasks.
Several points are noteworthy in this segment. The first is that task 3.3.5 (check for left-turning traffic) is duplicated from the previous segment, which occurs shortly before this one. The reason for this repetition is that in the previous segment this task was a decision criterion, whereas in the present segment it acts as a safety check for hazards. Another point is that whereas task 3.3.6 (accelerate to get through the intersection) is not part of the McKnight and Adams task analysis(9) and may not represent the safest action, it was included based on driver focus group findings involving similar intersection scenarios. In particular, these findings indicated that many younger and middle-aged drivers would be inclined to accelerate after they have decided to go on yellow to make sure they will get through before the light turns red.(21) Also, as discussed in the previous segment, drivers that are a little farther back when the light turns yellow may not have the option of either stopping safely or making it through the intersection legally unless they accelerate. Thus, this step was included because it likely reflects the manner in which many drivers act and also provides more of a worst-case scenario in terms of driver workload demands. Finally, it was assumed that the following vehicle would stop on the yellow light, so there is no need to maintain safe distance from the following vehicle.
Task Pacing and Timing - Tasks 3.3.1 (lane maintenance) and 3.3.6 (accelerate to get through intersection) are forced-paced because they are part of the ongoing task of driving. Tasks 3.3.4 (check for red-light-running traffic) and 3.3.5 (check for oncoming turning vehicles) are forced-paced because they have to be completed as the vehicle is rapidly approaching the intersection.
Regarding the task ordering, tasks 3.3.4 and 3.3.5 are ordered based on which one is encountered first. Also, task 3.3.6 (accelerate through intersection) is just a variant of task 3.3.3 (maintain safe distance from lead vehicle), and it is likely to follow the hazard-checking tasks.
To help identify potential information processing bottlenecks in this scenario, workload estimates from all the segments were combined into a single scenario-wide workload profile that provides a general indication of where the areas of high workload demands are likely to be.
Figure 27 shows the summed workload estimates (separately for each information processing subtask) in each segment interval for the entire scenario. Also, the intervals in which key tasks are forced-paced are shaded in orange. As indicated by figure 27, the workload peaks at higher levels for both the perceptual and cognitive subtasks during the Decision to Proceed segment. Perceptual workload also reaches moderate levels in both the Approach and Intersection Entry segments.
Figure 28, which displays the average workload estimate of all the tasks in play during a particular segment interval, shows that the peak during the Decision to Proceed segment in figure 27 arises from several tasks combined rather than fewer, more difficult tasks. This is because the average workload values in this segment are not much greater than in the other segments. Figure 28 does indicate that the peak for cognitive elements occurs in the Decision to Proceed segment, but the peaks for the perceptual subtasks mostly occur during the Approach segment.
Intervals containing nonroutine forced-paced tasks are shaded in orange. This graph shows the overall level of workload associated with a segment.
Figure 27. Scenario 3–Straight on Yellow Light total estimated workload ratings
Intervals containing nonroutine forced-paced tasks are shaded in orange. This graph generally represents the overall level of difficulty associated with the tasks in a segment.
Figure 28. Scenario 3–Straight on Yellow Light average estimated workload ratings
per task for each scenario segment.
Information about the combined and average workload ratings, pacing of key tasks, and nature of bottlenecks for each segment are shown in table 40. Only information that represents potential problems is listed; blank cells indicate that no substantive issues occurred for a particular segment or cell. Following the table is a list of key information processing bottlenecks identified in each of the segments.
Approach nature of bottleneck: Visual demands:
Decision to Proceed nature of bottleneck: High time pressure:
Intersection Entry nature of bottleneck: Concurrent moderate-to-high workload perceptual tasks with time pressure in the initial stages of the segment:
Topics: research, safety, intersection safety
Keywords: research, safety, task analysis, driver behavior, intersections, information processing, driver workload
TRT Terms: Motor vehicle drivers--Psychology, Human information processing, Traffic safety--Psychological aspects