Skip to contentUnited States Department of Transportation - Federal Highway Administration FHWA Home
Research Home
Report
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®

SECTION 4. SUMMARY AND CONCLUSIONS

Summary of Results

Table 91 shows a summary of the nature of the information processing bottlenecks for each scenario. This table helps identify recurring patterns among the segments, in addition to the key bottlenecks. The table's purpose was to integrate information across scenarios and segments to find consistencies on which to base more global conclusions. One caveat in interpreting table 96 is that there is redundancy within segments (columns) because many segments were comprised of the same set of core tasks with slight variations in the difficulty, number, or combination of additional tasks.

The cells in table 91 show bottleneck summaries. Scenario segments that did not have any significant bottlenecks are indicated with the text "None," and scenario and segment combinations that did not occur are indicated with the text "NA."

The following text gives a detailed summary of the key information processing bottlenecks encountered in each segment.

Approach Segments Summary of Key Information Processing Bottlenecks

Generally, the difficulties associated with the Approach segments involve moderate visual demands arising from information acquisition requirements. These demands include some combination of tasks, which involve determining if the light is about to change, identifying intersection characteristics, and identifying an unfamiliar intersection as the turn intersection. From an information processing perspective, workload demands in some of these tasks could be reduced by making it easier for drivers to perform these tasks.

One potential problem involves the subject driver having to identify an unfamiliar intersection as the turn intersection. The key difficulty with this task relates to the limited time window for completing this task. Specifically, the street sign does not become readable until the driver is sufficiently close to the intersection, and the information must be read and recognized before the driver is too close to the intersection to decelerate safely. Although this time window in the present analysis provided a relatively long duration to perform this task (2.8 sec), for several reasons this duration may still be limiting. In particular, the visual search component of the task can be made more difficult because of inconsistent placement of signs, a complex visual environment, or placement of signs along sight lines that are not visible until the subject vehicle is relatively close to the intersection. Similarly, a variety of factors affect sign legibility, including font characteristics, distance, and illumination/reflectance. Dingus et al., list factors that may increase the time needed to read the sign.(27) The difficulty and duration of this task not only affects the driver's ability to brake safely, but, more important, it also represents time that drivers are unable to spend completing other tasks, such as scanning the roadway for potential hazards. Potential strategies for addressing this problem include increasing the conspicuity and legibility of street signs.

Table 91. A summary of the nature of the information processing bottlenecks for each scenario.

Segment

Scenario

Approach

Prepare for
Lane Change

Execute
Lane Change

Deceleration/
Stop

Decision to
Proceed

Intersection
Entry

Prepare
for Turn

Execute Turn

1–Left on Green

Moderate visual demands.

NA**

NA

Several concurrent tasks.

NA

None

Concurrent high workload, high-stress tasks under time pressure.

Concurrent high workload, high-stress tasks under time pressure.

2–Left on Yellow

Moderate visual demands.

NA

NA

None.

High time pressure.

High time pressure for several tasks.

NA

Concurrent tasks.

3–Straight on Yellow

Moderate visual demands.

NA

NA

NA

High time pressure.

Concurrent perceptual tasks with time pressure.

NA

NA

4–Straight on Green

None.

High time pressure.

High time pressure.

NA

NA

Distributed information.

NA

NA

5–Right on Green

Moderate visual demands.

NA

NA

Several concurrent tasks.

NA

High time pressure for several tasks.

NA

Concurrent tasks.

6–Right on Red

None.

High time pressure.

High time pressure.

A relatively high number of concurrent tasks.

NA

NA

Concurrent high workload, high-stress tasks under time pressure.

Concurrent high workload, high-stress tasks under time pressure.

7–Stop on Red

Moderate visual demands.

NA

NA

A relatively high number of concurrent tasks.

NA

None.*

NA

NA

* This segment is "Proceed Through Intersection," but it is sufficiently similar to "Intersection Entry" to go in this column.

** NA–Not Applicable: This segment was not a part of this scenario.

Prepare for Lane Change and Execute Lane Change Segments Summary of Key Information Processing Bottlenecks

A general problem that arises in intersection-related lane changes is that they are more likely to be time limited compared to lane changes on highways. Intersection-related lane changes are usually more time limited than lane changes on highways because the approaching intersection acts as a forced completion point. Given that a key safety aspect of lane changes is checking for potential conflicts in the destination lane (e.g., mirrors and shoulder checking), the time limitations may cause some drivers to compromise these actions. This finding is also corroborated by the crash data.(21)

Another issue that was explicitly avoided in the current task analysis but which likely complicates the lane-change maneuver relates to whether a lane change is executed in conjunction with deceleration (i.e., when the speed of lead and following vehicles in the destination lane is also changing). More specifically, the lane change would be more difficult because of the need to accommodate the changing speeds of vehicles in the destination lanes. In addition, other drivers will have higher workload demands because they are approaching the intersection, which potentially diminishes their ability to respond to the subject vehicle's lane change and any potential conflicts.

General strategies for addressing lane-change-related safety issues might involve reducing the need to make lane changes, or making the information that is used to initiate a lane change available sooner in the approach to give drivers more time to change lanes. Note that one challenge that most strategies will have to face is that lane changes may be more variable with regard to when/where they are initiated; thus, it will be necessary to accommodate vehicles that are in a range of locations along the approach.

Deceleration/Stop Segments Summary of Key Information Processing Bottlenecks

The Deceleration and Stop segments were not typically associated with high workload; however, information processing bottlenecks that did occur tended to involve a high number of concurrent tasks. The key deceleration-specific tasks included decelerating at the necessary rate and maintaining appropriate spacing from other vehicles. These tasks represent ongoing activities that require repeated visual information acquisition and cognitive assessment of vehicle trajectories. In particularly complex situations with lead and following traffic, these visual and cognitive demands have the potential to interfere with drivers' ability to watch for and respond to other hazards in the roadway. Another potential problem involving decelerating and stopping relates to when the deceleration level is too low to permit gradual and safe slowing (e.g., intersections near freeway off ramps or on high-speed rural roads). Strategies for addressing these problems could focus on eliminating the conflicts between the subject vehicle and other vehicles and providing better information/cues about the subject vehicle's deceleration level.

Decision to Proceed Segments Summary of Key Information Processing Bottlenecks

The key difficulty with this segment appears to be a limited duration to conduct the tasks required to decide whether or not it is safe to proceed. For example, in Scenario 4, drivers had a time period of approximately 1 sec to make four different key judgments about decision-critical elements of the situation. Both the task analysis and focus group research conducted for this effort indicate that the decision to proceed is not based solely on the subject vehicle's position relative to a theoretical braking distance (although drivers must still determine this information). Instead, other information must be determined, which varies based on the complexity of the situation (e.g., presence of other vehicles). Moreover, obtaining this information involves some effort because the tasks are demanding and the required information is distributed throughout the visual environment. The task analysis shows that in complex situations, completing decision-critical tasks likely takes longer than the time available during the option zone, in which drivers still have the ability to both stop and proceed safely and legally. Strategies for addressing this problem might involve ensuring that the drivers' option zone is adequate to accommodate the necessary decisionmaking tasks. In standard calculations of the dilemma zone, the option zone is implicit in the driver's perception reaction time.(23,28)Thus, making this reaction time better reflect the complexity of different intersections in terms of configurations, traffic conditions, and other relevant characteristics may help address this issue.

Intersection Entry Segments Summary of Key Information Processing Bottlenecks

In general, the key difficulties in the Intersection Entry segments are likely associated with drivers having to quickly check for potential hazards (e.g., red-light runners, oncoming vehicles turning left.) at multiple distributed locations just before they reach the intersection). This is a greater problem, however, in scenarios in which the subject driver is about to make a turn without first coming to a complete stop. In this case, there are typically additional tasks (e.g., checking for hazards in the turn path) along with less time available to perform the tasks, because the vehicle is quickly approaching the turn location. In addition, interference between these tasks is likely because they draw on the same information processing resources. Strategies for addressing these problems could focus on reducing or eliminating the need for drivers to check for some of these types of hazards by either restricting the access of the hazards to the drivers' path or by providing notification of potential hazards.

Prepare for Turn Segments Summary of Key Information Processing Bottlenecks

The primary bottleneck in this segment is a difficult and forced-paced gap-judgment task that is complicated by having to quickly cycle among other tasks involving checking for hazards (e.g., in turn path, following vehicle) that are distributed throughout the visual environment. This is also likely to be a high-stress situation, because the consequences of making an error in the gap judgment task could be a collision with a fast-moving vehicle. Strategies for addressing these problems could focus on making the gap-judgment task easier to perform, reducing or eliminating the conflicts between the subject vehicle and other vehicles, or reducing the need for drivers to cycle their gaze among additional potential hazards.

Execute Turn Segments Summary of Key Information Processing Bottlenecks

The initial portion of the Execute Turn segment, in which the subject vehicle is in the process of turning, can be associated with higher levels of workload demands. The cognitive elements of this task are particularly affected because they generally involve overseeing precise maneuvers and assessing the situation for hazards. In addition, these tasks are forced-paced because the subject vehicle is quickly accelerating and, in some cases, directly in conflict with other vehicles. This is also likely to be a high-stress situation, because the consequences of making an error in executing the turn could be a collision with a fast-moving vehicle. The strategies for addressing these problems would likely be similar to those in the Prepare for Turn segments.

General Conclusions on the Nature of the Bottlenecks

At a basic level, the current analysis shows that information processing bottlenecks are likely to occur in specific situations during intersection driving. In addition, there appears to be a relatively high degree of consistency regarding the nature of bottlenecks across segments and scenarios. More specifically, time constraints and forced-pacing of tasks seem to be recurrent issues regarding information processing bottlenecks. The general problem seems to be that there is a limited amount of time available to perform a variety of different yet necessary tasks. Also, many of these tasks draw on the same perceptual and cognitive resources, which leads to more time-consuming sequential execution and also increases the potential for "interference" between tasks. This consistency related to the underlying causes of information processing bottlenecks makes it possible to apply this task analysis approach to identify potential safety issues.

Another general conclusion about bottlenecks is that in many cases it appears to be possible to develop strategies that specifically address the underlying information processing limitations associated with specific types of bottlenecks. For example, the current task analysis identified high visual/perceptual demands as a source of potential information processing bottlenecks during several of the Approach segments. The specific problems identified, including high information acquisition loads and time pressure in some instances, directly lead to strategies for addressing related problems (e.g., making those information acquisition tasks easier and quicker to perform). Thus, by gaining a more complete understanding of the underlying information processing aspects of potential problems, it is possible to identify general strategies that can directly address key underlying challenges associated with specific intersection driving situations.

Advantages of the Task Analysis Approach

The key advantage of the task analysis approach taken in this effort is that it provides unique information and perspective on intersection driving. In particular, this task analysis provides a useful complement to existing approaches for investigating driving behavior, such as crash data analyses, performance studies, focus group research, and even more traditional task analysis approaches. Set out below is a brief summary of some limitations of these approaches compared to the present task analysis approach:

  • Crash data analyses. These are focused on outcomes and provide limited information or understanding about driver actions or activities that lead to crashes.
  • Performance studies. These provide important information about driver activities and tasks, especially their dynamic aspects; however, they typically focus on a small range of driving situations and lack a common framework for synthesizing information across driving situations.
  • Focus group research. This approach is good for getting information about driver attitudes, beliefs, motivations, and the like, in various driving situations and it also useful for obtaining a global sense of what various driving situations entail. Focus groups are not good for getting specific information about driving tasks, however, and the information generated varies in detail, and often in quality, across situations.
  • Traditional task analyses. This method provides the same systematic approach to identifying driving tasks in different scenarios but it generally lacks the additional knowledge about the information processing aspects and scenario dynamics that is key to identifying potential bottlenecks.

The limitations discussed for the previous approaches demonstrate that the current approach provides valuable information not generally available through typical driving safety research approaches.

Another key advantage is that this current approach is well suited for providing information about the potential effectiveness of safety countermeasures. As shown in the previous discussion of the nature of segment-specific bottlenecks, this approach provides specific information about the underlying information processing elements associated with the bottlenecks that can occur in specific driving situations. Thus, this information can point to general strategies for addressing these potential bottlenecks and, more importantly, provide a basis for identifying countermeasures that may be well suited to addressing specific types of safety problems.

In addition to the advantages discussed previously, another benefit of the current approach is that it allows converging evidence to be used to identify potential bottlenecks. In particular, this approach employs indicators of overall workload, time constraints/scenario dynamics, identification of conflicting tasks (that draw on the same resources), and general distribution of information sources. These converging information sources allow researchers to attach some level of confidence to their conclusions about potential bottlenecks, in addition to providing a more complete understanding about the underlying nature of these problems.

Finally, the current task analysis approach is a relatively low-cost way to get this information. In this particular instance, existing task analysis research was used to populate the scenario task lists. In addition, the dynamics of the scenarios were estimated computationally using vehicle kinematics and prespecified assumptions about each scenario. All of these were analytical activities that could be conducted without requiring expensive data collection.

Future Research

There are several directions for future research involving the task analysis approach used in this research effort. The most obvious future research direction is to apply this approach to additional intersection or roadway configurations. This task analysis can be applied to most driving situations that take at least a few seconds to navigate and that involve sequences of actions that contain some degree of complexity. Following is a list of some examples:

  • Stop controlled and uncontrolled intersections.
  • Intersections with different configurations such as T-junctions and lane channelization.
  • Other roadway features such as roundabouts and complex multi-exit freeway exits.

One consideration when applying the current approach to other scenarios is the availability of crash data for potential scenarios, which would be useful for identifying candidate analysis situations and also for identifying specific scenario details–particularly if data exist on driver-related causal factors. Overall, these applications would provide the same type of information as the current analysis, which primarily includes information about potential information processing bottlenecks and insight into strategies for addressing the corresponding safety issues.

In addition to new scenario types, it is also possible to examine variations in existing scenarios to gain an understanding of how the task composition and overall scenario dynamics might be different in other situations. This examination would likely involve changes to key assumptions (e.g., travel speeds, sight/reading distances), which would in turn affect what tasks should be included and when they could occur. Possible variations could include some of the following examples:

  • Capacity-limited drivers such as older drivers.
  • Adverse weather conditions.
  • Intersections with different speed profiles such as rural intersections.
  • Intersections with different types of visibility issues such as street parking that might have sight obstructions caused by large vehicles, sight triangle obstructions such as trees, or intersections on grades.

Conducting task analyses on these variations could provide converging support for identified bottlenecks. In addition, this process could find situations that may not have been identified as particularly problematic in the current task analysis but could potentially become bottlenecks under less favorable conditions.

Another application may be the analysis of individual intersections, such as new or redesigned intersections that are in the planning stages. Using the task analysis in this way could provide some initial indication of potential safety issues in absence of crash histories or other operational data. One advantage of using the task analysis on a single intersection is that more detail could be incorporated into the assumptions (e.g., actual sight triangles, specifications of intersection configuration and elements), which would provide more precise timing and dynamics information, in addition to providing better guidance for the allocation of tasks within scenarios. With more development, the process of conducting a task analysis eventually could be automated or established with procedures to make the task analysis more cost effective.

Finally, another important area of future research could involve conducting empirical research to help validate the allocation of tasks within individual scenarios. For example, focus group research could be conducted to determine the extent to which drivers concur with the task lists and sequences, identify the location of bottlenecks, and identify which tasks drivers are more likely to skip when time is limited. Another important source of empirical validation could come from examining data from naturalistic driving studies, such as the 100-car study at Virginia Tech Transportation Institute.(29) These data could provide important information about the allocation and sequencing of observable tasks as well as the temporal dynamics of driver behavior. Questions about some assumptions made in the current task analysis could also be refined based on data from in-vehicle studies. For example, data are lacking on how far away from the intersection drivers begin making lane changes, and how much time and distance they take to complete the lane change. In general, empirical research could provide important guidance about overall driver strategies for navigating different driving situations as well as provide information about when and how frequently drivers perform certain key tasks.

Previous | Table of Contents | Next

ResearchFHWA
FHWA
United States Department of Transportation - Federal Highway Administration