The objective of this task was to define system requirements and test methodology for certain crash countermeasure systems. This task included identification of the relevant crash scenarios, development of vehicle level function specifications for the major countermeasure systems, and definition of test methodology for each countermeasure system. The expectation of this task was to assist in the development of the countermeasure systems by defining the crash scenarios that each countermeasure was expected to impact. A test protocol was expected so that each countermeasure system could be evaluated and its potential crash avoidance capability could be estimated. This task achieved definition of crash scenarios and a first attempt at a test protocol definition.
Relevant crash scenarios were identified for the forward, side and rear direction of vehicle travel. The frequency of crashes, as well as estimates of annual losses due to direct cost and years of function life lost was tabulated (see Figure 3.1 below). For more information on how the tabulations were developed refer to Ted Miller's paper, Understanding the Losses from US Motor Vehicle Crashes (AAAM, 1995). Currently the percentages represent estimates of the total crash avoidance opportunity and were not adjusted for a particular technology.
An important development of this study involved an analysis of crash data. By examining a number of state data bases for police reported crash data, information for over represented conditions can be developed. This was especially important when assessing potential countermeasure impact. Conditions of wet or icy road surface, alignment of the roadway, and type of road can be studied for level of involvement and whether the condition was over represented when compared to other crash types. If some condition was shown to be over represented, it was not assumed to be the cause of the crash but was included in the scenario development. If no condition was over represented, the crash has a probability model based on the distribution of all crashes.
In the forward direction, rear end crashes (same direction of travel) were the primary crash configuration. For side direction, lane changes, merge and lane departures (all same direction of travel) were the configurations of interest. For the rear direction, backing crashes were studied. An example of a crash description as defined for this task is shown in Figure 3.2. The one page description includes a defined heuristic crash with any identified over represented condition, an estimate of the direct cost realized annually, and an estimate of years lost of functional life per year.
Figure 3.1: Estimate of Annual Losses due to Automotive Crashes.
Crash #62: Inattentive, Rear
SCENARIO: A northbound vehicle, A, is stopped waiting at a red traffic signal in an urban area on a major artery. Another vehicle, B, coming from some distance behind, doesn't notice that A is stopped and cannot stop in time. (No other conditions have yet been identified as over-represented in this crash.)
KEY STATISTICS: In Focus Group Interviews respondents indicate driver B was usually looking away. Also per these same interviews, many of these go unreported.
LOSS INDEX: Vehicle Crashed 12.0%
Percent Direct Cost Per Year 10.2%
Percent Functional Years Lost 4.9%
REFERENCES: From the union of the Indiana Tri-Level Causation study nine percent of all crashes involve “Driver: Inattention,” and from the University of Michigan Transportation Research Institute (UMTRI) Michigan crash typology 14% are “Multiple Vehicle: Rear End,” Office of Crash Avoidance Research (OCAR) found 24.2% of all crashes to be rear end crashes. Of these 70% (17% of all crashes) involved a stopped lead vehicle and therefore do not involve “coupled” vehicles. Of all the crash causes of the striking vehicle (10%), “driver inattention to the driving task” is the most common error. (Proceedings of IVHS America 1993 Annual Meeting, P.251).
Figure 3.2: An Example of a Crash Description.
Also, a time line for a sequence of events introduced the concept of “Time to Collision” (TTC) for forward defined heuristic scenarios. Parameter values were identified for range, range rate, vehicle attitude, host velocity, host deceleration, host vehicle response time, and driver response time. Some of these values can be measured and some can be calculated based on the particular countermeasure capability. TTC was defined as the range divided by the range rate with the units of seconds. The measure of parameter improvement due to a given countermeasure was difficult to characterize. A countermeasure will eliminate or mitigate a crash when it provides an opportunity for the driver to avoid involvement in a crash by improving driver response time allowing the driver to remain in control. This can also be stated as the sum of the driver reaction time and the required stopping distance time or avoidance time (given the road conditions and capability of the vehicle) is less than the calculated TTC.
In summary, the target crashes were identified and sized and were referred to as the relevant scenarios. The relevant crash scenarios of rear-end crashes are shown in Table 3.1. In an effort to provide an alert during the relevant crashes, the system design may be vulnerable to nuisance alerts. Development scenarios require representation of a cluttered background, competing obstacles, and geometric concerns of roadway alignment to address nuisance alerts. These development scenarios, although not necessarily statistically high in crash frequency, need to be evaluated through the testing process to evaluate system performance.
The purpose of this task was to establish user requirements based on quantitative evaluation of the user's ability and the environment in which the system was designed to provide the driver with an opportunity to avoid involvement in forward, side and backing crashes. This was done without reference to a specific countermeasure and no actual system design parameters were discussed. User requirements for four systems were addressed: forward collision warning, lane sensing, side near object detection/warning, and rear near object detection/warning systems.
The area of coverage for forward collision warning (Figures 3.3 and 3.4) was based on vehicle kinematics. The extent of the forward range was a function of the relative speed between the host and identified collision object. The limitations on height and width of field reduce nuisance alerts but provide lane coverage under transitions in road alignment. Increasing the width of coverage to include same and next lane (right and left sides) will allow for collision assessment based on lateral motion in the forward traffic field. Providing the driver with improved response time may reduce rear end collisions and lane departure due to unplanned avoidance maneuvers.
Table 3.1: Scenarios for Requirements
|
|
|
|
Frequency | Potential* |
| Forward Object Stopped |
|
Maximum longitudinal range. Path accuracy to sort primary target from other stationary clutter (height and width of anticipated path). Latency in new target acquisition time. |
|
|
| Forward Object Moving |
|
Target separation to identify primary target in traffic (motorcycle in lane with a large truck in the next lane). Latency in new target acquisition time and coasting of target no longer present. Maximum negative range rate for reassignment of primary target. |
|
|
| Tailgating |
|
Range rate threshold response |
|
|
| Cut-in |
|
Minimum longitudinal range |
|
|
| Head On |
|
Excess of maximum range rate (Maximum: host vehicle to stopped object) |
|
|
| Pedestrian |
|
Minimum size object for detection |
|
|
| Weather |
|
System performance under low visibility conditions and / or low coefficient of friction. |
|
|
*Potential: High potential implies that the countermeasure has a 20% probability of assisting the driver. Medium implies 10% probability and Low implies 5% probability.
Lane sensing augments the function of forward collision warning with path prediction and collision assessment. There may be other benefits such as identifying lane and road departures, but no effort was made to explore this function.
The area of coverage for side detection is shown in Figures 3.5 and 3.6. The rear side zone aids the host driver in lane change maneuvers. The forward side zone function augments collision possibility involving the other driver's intended and the host's unintended lane changes for the forward area of coverage.
Figure 3.3: Forward Detection Zone
Figure 3.4: Front View Forward Area of Coverage
Figure 3.5: Side Near Object Detection Area of Coverage (Plan View)
Figure 3.6: Side Near Object Detection Area of Coverage (Elevation)
The area of coverage for the rear of the vehicle (Figures 3.7 and 3.8) should assist the driver in detecting near stationary and slow moving objects. The system is only active when the transmission is shifted in reverse. The coverage zone shows a radius of 5 meters. The actual value should be representative of the turning radius of the host vehicle. An additional function of a rear detection system is as a parking aid. This requires multiple zone detection or high accuracy at close range.
The areas defined were based on both the statement of the crash scenario, vehicle kinematics, and the rational as shown in Figure 3.9. One requirement of the technology was recognition of stopped in-path objects on the roadway. Making the distinction between “in path” on the road and “threat” drove the requirements for update rate, accuracy, field of view and range.
The goal of this task was to define a test methodology at the Driver-Vehicle System level for the countermeasure system investigated. The deliverable was a summary report including a test protocol designed to assess the crash reduction potential for these countermeasures. Although the original scope of the task was only to develop this test protocol, a near term test protocol proposal was added to the driver-vehicle system test protocol. The driver-vehicle system proposal did address the statement of work but it required an investment of significant resources to develop and implement. While this test protocol has long term potential, this protocol could not be accomplished within the scope of this project.
Three concepts were stressed: (1) Testing at this level should include a high level metric that was simple to measure, (2) The tests should measure the effect a countermeasure has on the relevant crash scenarios, and (3) The near term tests can only estimate the maximum potential reduction of crashes due to a countermeasure.
The metric developed for this project was Time to Collision (TTC) and Time to Avoidance (TTA). The relative range between two vehicles divided by the range rate is TTC, and the characteristic width of the vehicles divided by the transverse velocity vector perpendicular to the range rate vector is TTA. If TTC remains less than TTA then a collision occurs. However, if TTA becomes less than TTC as a scenario proceeds, then the collision is avoided.
Figure 3.7: Rear Detection Zone (Top View)
Figure 3.8: Rear Detection Zone (Side View)
Figure 3.9: Vehicle Interaction Definition and Formulation
This metric was chosen for its ability to measure the proximity of a crash between two vehicles. It is independent of the crash scenario, the avoidance maneuver chosen by the driver, and the countermeasure used. The raw data required is a time history of the relative locations of the vehicles. It can be used in simulator and closed course tests.
The Near-Term Tests defined in this task provided a test plan for evaluation of the ACAS collision avoidance systems, with emphasis placed on the near-term testing in ACAS Task 6.4. The tests outlined in these sections would be conducted by trained drivers, who would be aware of the test scenario scripts.
The tests are designed to measure the TTC and TTA when the system provides an alert. Because the driver performance will be controlled due to this experimental design, several effects that a given countermeasure system may have on driver performance will not be assessed. These effects include those of nuisance or false alarms and early alerts. Although this phenomenon will be noted during testing, no determination of their effect on driver behavior will be made.
Due to these limitations, the results from these tests would only provide a measure to estimate the maximum countermeasure potential of crash avoidance for the scenarios tested. Because these tests will include a sub-set of the scenarios outlined earlier in this document, the maximum potential crash avoidance would be based on only the tested scenarios.
It is not possible to test a collision avoidance system under all possible combinations of potential performance limiting factors. Instead, it is necessary to devise a reasonably sized series of tests which subjects the system to factors sufficient to infer system performance under untested conditions. The following discussion describes a possible test protocol for near-term tests.
A fractional factorial design will minimize the number of tests required for near term evaluation. This design will provide information on the effect of measured parameters. The purpose of these tests is to determine the TTA and TTC at the point when the system provides a driver with an alarm. The variables chosen will test the system's ability to provide an alarm in situations found in the driving environment. The fractional design allows for an estimate of the results of these tests regardless of whether or not they are caused by a measured parameter or an interaction of parameters.
These near term tests would consist of one experiment for each of the countermeasures being evaluated. These experiments would all include the same output variable of TTA and TTC at the point of warning. They would also include the same four control variables.
The first control variable is the test scenario. There are three scenarios for each countermeasure. The second control variable is the cooperative vehicle size. The “levels” for this variable are a medium duty truck, a medium sized sedan, and a motorcycle. The overhead view of these three potential cooperative test vehicles is shown in Figure 3.10. For the rear near object detection systems (NODS) experiment, the medium duty truck will be replaced with a child's tricycle.
Figure 3.10: Overhead View of Potential Cooperative Test Vehicles
The third control variable is the initial relative velocity between the test vehicle and the cooperative vehicle. This variable will have three values: 25, 35, and 45 miles per hour. The velocities are listed in the descriptions of each experimental design and are dependent upon the countermeasure evaluated.
The final variable is the level of clutter. The values of clutter will be none, rural, and urban. The physical layouts of these clutter levels will be well controlled and documented in a map known as the ground reference. A clutter map will be created for each of these levels. The clutter levels are listed in the descriptions of each experimental design and are dependent upon the countermeasure evaluated.
The requirements and test protocol provide a basis for future development of test protocols to evaluate the potential crash avoidance capability of countermeasure systems. The Crash Avoidance Metric Partnership (CAMP) agreement with NHTSA should provide test procedures for forward collision warning countermeasure systems. The system requirements and crash scenarios defined in this task have provided the CAMP project an initial start in their effort to determine test protocols.