U.S. Department of Transportation
Federal Highway Administration
1200 New Jersey Avenue, SE
Washington, DC 20590
202-366-4000
Federal Highway Administration Research and Technology
Coordinating, Developing, and Delivering Highway Transportation Innovations
| SUMMARY REPORT |
| This summary report is an archived publication and may contain dated technical, contact, and link information |
|
 |
| Publication Number: FHWA-HRT-17-025 Date: December 2017 |
Publication Number: FHWA-HRT-17-025 Date: December 2017 |
This experiment examined driver performance as a function of set speed, assigned gap-setting, and preferred following distance.
At 65 mi/h, a 1-s gap would leave approximately 95 ft between vehicles. Previous studies have shown that drivers felt both comfortable and safe traveling with gaps shorter than 1 s. For example, in an on-road study testing drivers’ choices in following distances, drivers regularly used gap settings shorter than 1 s; when following another vehicle, drivers elected to set the gap at 0.7 s or 0.6 s 80 percent of the time.(16) However, with a 0.6-s gap, there is approximately 57 ft between vehicles. If an average vehicle length was assumed to be about 17 ft, this would leave less than 20 ft of buffer in the front and rear for a merging vehicle. As a result, at these shorter distances, drivers might not feel comfortable merging or having a vehicle merge in front of them. Furthermore, gap-based discomfort might vary from person to person.
It is possible that individual differences in preferred following time gap might be related to perceived driver workload or driver capabilities to react in events that require the driver to override the CACC system and resume manual control.
This experiment was conducted in two parts. The goal of the first part was to estimate the median preferred following distance of drivers in the participant pool. This estimate was then used in the second part to assign new participants to groups based on their preference for far (i.e., longer than the median preferred following distance) or near (i.e., shorter than the median preferred following distance) car following distance.
The second part was a near replication of CACC with the crash event condition described in chapter 2. The differences between the CACC with crash event scenario described in that chapter were the following:
The goals of the second part were the following:
The following subsections describe the method, equipment, scenario, participants, and procedure for the part 1 experiment.
This section describes the equipment, simulation scenario, participants, and procedures for part 1 of the experiment.
A National Advanced Driving Simulator (NADS) ¼ cab MiniSim™ was used. Three 42-inch 720‑pixel plasma screens displayed the forward roadway, side, and rearview mirrors. An additional 12-inch widescreen displayed dashboard information. The simulator was fix based and used a subwoofer beneath the driver’s seat to generate road feel. As in the preceding experiments (see chapter 2), the simulated vehicles were scaled to 75 percent of the original scale. Several pilot subjects were used to verify that the down-scaling of the lead vehicles was appropriate for this simulator.
As in the previous experiments, participants drove in a dedicated center lane on a simulated eight-lane interstate highway. Entrance to the dedicated lane was accessed from a left-side ramp. The dedicated lane was separated from other lanes with a jersey barrier. The environment was similar to suburban-rural interstate driving with a mix of trees and buildings along the roadway.
The simulation began with the participant vehicle as the only vehicle on the roadway. During this time period, participants practiced steering and braking to accustom themselves to the feel of the simulated vehicle. After a few minutes, participants came upon another vehicle. That vehicle acted as a lead vehicle and drove at 55, 65, 70, and 55 mi/h for 3 min each.
Participants were 14 licensed drivers recruited from the Washington, DC, metropolitan area. Participants were required to be at least 18 years of age and were screened for susceptibility to motion and simulator sickness. Half of the participants were male. Age ranged from 22 years to 72 years with a mean age of 46.7 years (median 50.5 years).
All participants completed two drives. The goal of the first drive was to assess the participants’ comfortable driving distance. Specifically, participants were asked to “ drive at what you consider a comfortable distance. In other words, follow that vehicle at a distance that you would normally follow another car in the real world.” Participants were reminded that the lead vehicle would change speed several times and that speed would need to be adjusted to maintain an appropriate following distance. The lead vehicle drove at 55, 65, 70, and 55 mi/h for 3 min each. The first drive lasted between 14 and 17 min.
On completion of the first drive, participants completed a simulator sickness questionnaire.
The goal of the second drive was to assess drivers’ perceived minimum safe following distance. This drive was identical to the first except for the instructions. Participants were told the following:
…instead of following at a comfortable distance, I want you to drive more closely. I’d like you to follow that vehicle at the minimum distance that you might ever follow another car on the roadway. For example, imagine that you are on a busy road and are trying to change lanes. Or even if you were simply in a hurry to get somewhere.
Participants were given an opportunity to ask questions to ensure that the task was fully understood.
Each participant in part 2 performed three drives. The first drive was used to determine whether the participant preferred a near or far gap. Driver participants were given the same minimum safe gap choice instructions that were used in part 1. Based on this first drive, participants were assigned to one of two groups. They were assigned to the near preferred gap group if they drove with a minimally safe gap less than 0.9 s in their first drive. Participants who drove with minimally safe gaps greater than 0.9 s in their first drive were assigned to the far preference group.
The second drive was same as the first, but with the comfortable gap instruction to “…drive at what you consider a comfortable distance. In other words, follow that vehicle at a distance that you would normally follow another car in the real world.” The second drive was considered practice, and the data from that drive were not used in determining gap preference.
For their third drive, half of those who preferred a near gap in the first drive were assigned to drive with a near (0.6-s) gap, and half to drive with a far (1.1 s) gap. Likewise, half of the participants who preferred a far gap in the first drive were assigned to drive with a near gap and half to drive with a far gap.
This section describes the equipment, participants, and procedure used in part 2.
The same MiniSim™ simulator was used for part 2 as was used for part 1.
The same simulated eight-lane interstate highway was used in both parts 1 and 2. Participants in part 2 drove the same drives as the participants in part 1. In addition to these two drives, participants completed a third drive. The third scenario used the same roadway as the first two drives.
Participants were 99 licensed drivers from the Washington, DC, metropolitan area. Of these, 57 were classified as preferring a near gap and 36 as preferring a far gap. Each gap preference group had approximately the same number of males and females and approximately equal numbers of participants over and under 45 years of age. Approximately half of the participants in each preference, gender, and age classification were assigned to the near gap setting and half to the far gap setting.
Throughout the first two drives, the procedures were the same as for part 1. After the second drive, participants were briefed on the NASA-TLX and the CACC concept. Participants then moved back to the simulator where they were shown how to use the multifunction display to engage CACC, adjust the set speed, and set the gap distance. The multifunction display was the same as that used in CACC experiments 1, 2, and 3 described in chapters 2, 3, and 4 of this report, respectively. Participants then completed the third drive with the far or near gap setting that they had been assigned.
The third drive was modeled after the drive in experiment 3. The drive began with the participant stopped on an onramp in the third position of a four-car string. When the ramp meter turned green, the string proceeded down the ramp and accelerated to 70 mi/h while maintaining the appropriate gap (0.6 or 1.1 s). Approximately 5 min into the drive, another CACC vehicle merged in front of the participant halfway (30.6 ft or 56 ft) between the participant’s vehicle and the vehicle the participant had been following. The CACC system then adjusted the speeds of the affected vehicles to restore the assigned gap. If a participant braked in this situation, the CACC system disengaged and then needed to be reengaged.
Approximately 20 min into the drive, a vehicle sped down an onramp, merged in front of the platoon, and crashed. The crash was not in the participant’s line of sight. The crash avoidance event began when the lead vehicle in the string decelerated at 32 ft/s2 in response to the crash. One-tenth of a second after the lead vehicle began braking, all of the CACC vehicles behind it simultaneously began to decelerate at 0.4 g (12.8 ft/s2) and activated their brake lights. The simulator’s engine noise was configured to exaggerate the change due to the 0.4-g deceleration.
The following subsections describe the results of part 1 of the experiment.
The goal of part 1 was to estimate the median for perceived minimum safe following distance. This information was used to determine whether participants in part 2 would be labeled as far or near followers.
To provide participants with sufficient time to adjust following gap for each speed change (55, 65, 70, and 55 mi/h), the first 30 s of vehicle following at each speed were excluded from analysis.
Table 10 presents drivers’ following time gap distributions by speed averaged across 13 participants during the near following distance drive.
The mean median following distance time gap 0.91 s was used to assign participants in part 2 to far- and near-follower groups.
Participants in part 2 completed the same two drives as participants in part 1. The data from the first comfortable following task were not used to determine following distance preference. To provide participants with sufficient time to adjust following gap for each speed change, the first 30 s of vehicle following at each speed were excluded from analysis.
Data from the second drive were used to assign drivers to far and near following groups, with drivers with a mean following distance less than 0.9 s assigned to the near group. Table 11 presents the gap distribution as a function of speed under the near following distance instruction.
The NASA-TLX was administered verbally at three points during the third drive—shortly after the first merge, during a cruise period, and after the final crash event. The effects of preferred following distance and CACC set gap on workload were tested using a GEE with NASA-TLX measurement location as a repeated measure.
Figure 17 shows that as a result of workload being rated substantially higher following the crash avoidance event, perceived workload varied significantly with measurement location (χ2(2) = 129.81, p < 0.001). In addition, there was a significant three-way interaction (χ2(6) = 27.36, p < 0.001) that apparently resulted, because in their post-crash ratings, drivers assigned to the near gap reversed the otherwise consistent trend for drivers who preferred a near gap to rate workload lower than those who preferred a far gap.
Note: Error bars represent estimated 95-percent confidence limits of the means.
Figure 17. Graph. NASA-TLX rating as a function of measurement location, preferred gap, and gap setting.
The CACC system was programmed such that it was not necessary to override by manually braking during the merge event that occurred 5 min into the drive. However, the high speed and short distance between the participant and merging vehicles, especially in the near following case, provided a measure of driver trust or comfort. Only 3 of 99 participants depressed the brake pedal during or after the vehicle merge.
Another indication of trust, or caution, was whether participants hovered their foot above the brake during the merge event. Participants in the near following distance condition were significantly more likely to hover over the brake pedal than those at the far distance (χ2 (1) = 5.27, p = 0.022). Neither preferred following distance nor its interaction with other variables was statistically significant. The data suggest that following distance preference did not affect trust or caution during the cut-in merge.
Participant reaction time was calculated as the time between when the lead vehicle in the string began decelerating and when the participant initiated brake pedal depression. The results shown in figure 18 are based on data from 73 participants. Data files for six participants were corrupt and could not be read. A simulator failure caused loss of data for one participant. Three participants did not brake at all in response to the crash, and 11 participants were not included in the analysis because they were deemed outliers in that they did not initiate braking within 7.1 s, which represented more than 1.5 times the interquartile range (a standard definition of an outlier).(17) Participants who drove with the near gap-setting depressed the brake pedal significantly sooner than the participants who drove with the far gap-setting (χ2 (1) = 4.28, p < 0.04). No difference in reaction time based on preferred following distance was found, nor was the interaction between preferred and assigned following distance significant.
Note: Error bars represent estimated 95-percent confidence limits of the means.
Figure 18. Graph. Brake onset reaction time as a function of preferred and assigned gap.
Participants with the near gap setting had a crash probability of 0.82, whereas those with the far gap setting had a crash probability of 0.61. The difference in probabilities was statistically significant (χ2 (1) = 4.32, p = 0.038). No other significant effects were found.
As in experiment 1, reported in chapter 2, workload was rated higher after the crash event than before, where workload was rated fairly low. Interestingly, preferred and assigned following distance did not affect perceived workload during uneventful driving. Although failures to reject the null hypothesis (i.e., no difference in workload between near and far preference groups) must be viewed with caution, it appears from these results that using a cruise control gap setting less than the preferred following distance does not appreciably affect workload.
Reaction times to the crash event were shorter with the 0.6-s (near) gap than with the 1.1-s (far) gap. This finding should be expected because drivers have more time to react with the longer gap setting. The finding is consistent with cruise control being a convenience feature (i.e., that drivers delay braking to avoid relinquishing that convenience). Also, because there was no warning tone when the 1-g lead car deceleration began, drivers with the longer gap had no immediate reason to perceive a need to quickly intervene.
The reaction time difference between the CACC-B group (which also did not receive an audible alarm) in chapter 4 and the far gap group in this experiment were nearly identical at just over 3 s. This suggests that for car following experiments the higher fidelity moving-base simulator and the lower fidelity fixed-base simulator yield similar results. Based on the results presented in chapters 2 and 4, it is likely that an auditory alarm in this experiment would have reduced the crash rate when the following distance was 0.6 s; however, the extent of that reduction cannot be estimated with the data at hand.
