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 |
The viability of CACC as a successful and widely used technology is dependent on many factors. One of these factors is the ability of drivers to enter and exit strings of closely spaced CACC vehicles. It has been hypothesized that in the early stages of CACC market penetration, there will be a dedicated CACC lane.(6) The lanes would function much like high-occupancy vehicle or managed lanes. Drivers would only be allowed to travel in the dedicated lane if certain requirements were met. In this case, the drivers would have to be using CACC-equipped vehicles. These lanes would presumably have some sort of physical separation from the regular flow of traffic. This would prevent the disruption of the CACC flow due to non-CACC-equipped vehicles attempting to enter the traffic stream.
This chapter describes an experiment that explored the ability of drivers to enter and exit CACC strings in a dedicated lane. The goal of this research was to address human factors issues. Specifically, the goals of this experiment were to (1) investigate drivers’ abilities to successfully enter a dedicated CACC lane and join an already established vehicle string and (2) assess the workload associated with this maneuver.
CACC can be implemented in many ways. This experiment was based on several critical assumptions concerning vehicle technology and roadway infrastructure. These assumptions did not imply that the CACC system would ultimately be implemented in exactly this manner. Rather, they served as points of reference for addressing human factors issues.
There are several ways in which a driver could enter a CACC string in a location other than the front or rear (i.e., as the first car or the last vehicle in a string). One way would involve requesting permission to move into the platoon. In this case, the driver would request permission, permission would be granted, a larger gap would be provided between two of the vehicles in the existing string, and the driver would merge into the string. This method would require more complexity in the CACC operating system and driver interface than other methods. A second way in which a driver could enter a CACC platoon would be one in which vehicles did not request entry into a string, and an extra gap for those vehicles would not be created. Instead, the driver would have to merge into the platoon, and the other vehicles in the platoon would adjust speed to restore the desired gap and to accommodate the new platoon member. This method would result in a variety of human factors issues, especially when vehicles were traveling with short gap distances.
A third way to accomplish a merge into a CACC string would be to enable CACC longitudinal control of the merging vehicle during the merge itself. 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. In fact, overall, 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 would be only 57 ft between vehicles. If the average vehicle length was assumed to be 20 ft, this would leave less than 18.5 ft of buffer in the front or rear for a merging vehicle. As a result, drivers might not feel comfortable or have the skill to merge into a string without longitudinal assistance. Similarly, drivers might not feel comfortable allowing the system to assume longitudinal control during such a merge. For this reason, driver acceptance of longitudinal acceleration by the CACC system to merge was explored.
This experiment explored the following three different types of merges:
Three participant groups drove the same stretch of simulated limited-access roadway used in experiment 1. All groups were asked to exit and enter the roadway four times. The first group completed this task with no cruise control. The second group used CACC while traveling in the main roadway stream but was required to adjust the vehicle speed when entering and exiting the CACC stream. The third group was provided with CACC that controlled speed both while in the travel lane and while merging and exiting the CACC string.
The NASA-TLX was administered four times during the drive. The first assessment was during a practice drive to accustom participants to providing verbal responses to the workload protocol. The second workload administration was 5.8 mi into the drive, immediately after the first complete exit and reentrance into the CACC string. This administration assessed subjective workload imposed by the merge. The third assessment was 20 mi into the drive and assessed subjective workload associated with driving in a stable, unchanging state (i.e., a baseline index). At this point, drivers were between merging events and were likely to feel comfortable with the driving task. The fourth administration was immediately after the final merging event, approximately 23 mi into the drive. This final workload assessment was intended to assess the change in subjective workload that might have occurred following successive string exits and merges.
Physiological arousal was assessed by measuring eyelid closure, pupil diameter, and skin conductance. These measures were assessed at eight different 15-s periods during the drive. Four of these periods were immediately before exiting the roadway (i.e., cruise periods), and four were in the last portion of the merge events.
At exits 4, 8, 12, and 16, participants were asked to leave the CACC platoon by using a left ramp and then reenter the traffic stream using a left onramp. The exits were approximately 1.45 mi apart. The CACC platoon movement was continuous and did not stop, which forced participants to enter mid-string. The two groups with CACC were required to join the CACC string in which each vehicle maintained a 1.1-s gap. The group with CACC merge assist was not required to adjust speed in any way. The CACC system maintained longitudinal control during the entire drive unless the participant pressed the brake. The group without CACC merge assist was required to manage and adjust its own vehicle speed while merging into a 1.1-s gap. The third group drove without CACC and maintained longitudinal control at all times. This group was provided with a variety of gap sizes to merge into, which, it was hoped, would help determine whether participants generally preferred shorter or longer gap distances or showed no preference (i.e., drivers would accept any gap).
The following subsections describe the equipment and materials used for this experiment.
As with experiment 1, this experiment was conducted in the FHWA Highway Driving Simulator. For this experiment, the simulator’s motion base was enabled. Typical motion for roll, pitch, and yaw fell within ± 4 degrees.
The same eye tracking system was used as in experiment 1. Because merges were from the left, the right-side blind spot area was included as an area of interest rather than being included as part of the road-ahead as in experiment 1.
As in experiment 1, GSR was measured with silver-chloride salt electrodes placed on the palmar-side base of two fingers on the participant’s left hand. The electrodes were connected to a small sensor with a Bluetooth® transmitter strapped to the left wrist.
The roadway was the same one used in experiment 1 with a few minor variations. The entrance to the center dedicated lane was accessed from the left side of the roadway from a ramp. The simulation began with the participant’s vehicle as the third in the CACC platoon queue. Once the participant was ready to begin, the two vehicles in front of the participant accelerated and merged into the CACC lane and cruised at 70 mi/h. The two groups that drove with CACC engaged the system, and the participant’s vehicle maintained a 1.1-s gap between it and the vehicle in front of it. The control group participants could follow at any distance they chose.
There were 18 exit ramp/entrance ramp pairs, each placed approximately 1.45 mi apart. Participants were asked to use exits 4, 8, 12, and 16. Exit ramps to the left of the main travel path that were not used by the participants were blocked by traffic barrels. This was intended to serve as a reminder to participants of which exit ramps to use.
As in experiment 1, the CACC vehicles were scaled to be 75 percent of the actual size of the model of vehicle represented so that participants could more accurately perceive following distance.
Participants in all three experimental conditions were told the following:
I am going to ask you to exit and reenter the freeway every fourth exit. I will give you verbal reminders to exit the freeway. There will be orange construction barrels blocking the other exit ramps. There will be other traffic on the freeway. The traffic is continuous and will not stop.
Participants were given the following additional condition-specific instructions.
The CACC with merge assist instructions were as follows:
The CACC without merge assist instructions were as follows:
The control condition instructions were as follows:
The experimental scenario began with the participant seated in the third vehicle of a platoon of four vehicles. Once the participant was ready to begin, the two vehicles in front of the participant accelerated and merged into the CACC lane and cruised at 70 mi/h. The two groups that drove with CACC engaged the system, and the participant’s vehicle maintained a 1.1‑s gap between it and the vehicle in front of it. Participants in the control condition were asked to drive as they normally would, with no specific instructions given about following distance.
Participants were verbally reminded to exit the travel lane and then reenter traffic at the appropriate ramps. As soon as the participants successfully merged into traffic in the dedicated lanes after the first (exit number 4) and fourth (exit number 16) ramps, the NASA-TLX was administered to assess workload during the merge event (during the preceding minute or so). The NASA-TLX was also administered as soon as exit number 14 was passed. (Participants did not use this exit.) This administration was intended to assess workload during uneventful cruising in a CACC platoon (also described as during the last minute or so).
The primary between-group independent variable was the level of cruise control automation used throughout the scenario.
There were the following three distinct participant groups:
In addition to workload, there was one additional within-subjects variable, labeled period, with eight levels that were intended to distinguish the effects of CACC on driver behavior. The eight periods are described in table 3.
Usable data were obtained from 48 participants. Participants were required to be at least 18 years of age and were screened for susceptibility to motion and simulator sickness. Table 4 shows the age group and gender counts by treatment group. The mean age of the younger participants was 33.4 years (range 19.4 to 44.5 years). The mean age of the older participants was 56.6 years (range 46.5 to 77.9 years).
Participants were paid $80 for between 1.5 and 2 h of participation.
The following subsections describe the results of the experiment for workload, physiological arousal, distraction, merge behavior, steering entropy, visual behavior, and trust in the CACC system.
The NASA-TLX was administered verbally at three points during the experiment: shortly after the first merge (exit 4), roughly halfway between the third and fourth merges (exit 14), and shortly after the fourth merge (exit 16). Mean workload estimates obtained using GEEs are shown in figure 9. The location by condition interaction was significant (p < 0.02). The control condition had a significantly greater mean NASA-TLX score than the CACC with merge assist group and CACC without merge assist group (all p< 0.05). The interaction resulted because at exit 4, the CACC without merge assist group had significantly greater workload scores than the CACC with merge assist group, whereas this difference did not surface at exits 14 and 16.
Note: Error bars represent estimated 95-percent confidence limits of the means.
MA = Merge assist.
Figure 9. Graph. Estimated mean workload (NASA-TLX) by treatment group and location.
As with experiment 1, the physiological measures of arousal were GSR, eyelid opening, and pupil diameter.
GSR is generally considered to be sensitive to sympathetic nervous system arousal, and it is more sensitive to spikes in arousal than to gradual changes in arousal for longer periods of time. If merging were indeed stressful, higher levels of GSR should be seen for the merging periods (2, 4, 6, and 8) compared with the cruising periods (1, 3, 5, and 7). Furthermore, the drivers in the control condition should also exhibit greater levels of GSR compared with those who used the CACC system as a result of the arousal-reducing effect of the automation.
Mean-standardized GSR scores were analyzed using GEEs. Resulting mean estimates with 95‑percent confidence intervals for each condition and period are shown in figure 10.
Note: Error bars represent estimated 95-percent confidence limits of the means.
Figure 10. Graph. Estimated mean GSR conductance z-score by period.
As shown in figure 10, overall, GSR was significantly greater during post-merge periods than pre-merge periods. In other words, participants were more aroused during the post-merge periods than during the pre-merge periods.
However, participant condition did not significantly affect mean GSR values (χ2(2) = 1.49, p > 0.05). That is, the use of CACC did not significantly reduce arousal as assessed by GSR nor did time period interact with condition.
As people become more relaxed or tired, eyelids tend to droop. If CACC reduced alertness, one might expect eyelid opening to be smaller over time. As with GSR, the raw eyelid-opening measures were converted to z-scores. Eyelid-opening observations that the eye-tracking software classified with a quality rating less than 75 percent were excluded. This quality rating resulted in the exclusion of 39 percent of the eyelid-opening readings.
Neither experimental condition nor time period significantly influenced eyelid opening. The interaction between time period and condition was also insignificant.
Pupil diameter measurements for which the eye-tracking system reported less than 75 percent quality were excluded from the analysis. As with GSR and eyelid opening, each participant’s pupil diameter observations across the eight 15-s periods were converted to z-scores. Mean estimates with 95-percent confidence intervals for each condition and period are shown in figure 11.
Note: Error bars represent estimated 95-percent confidence limits of the means.
Figure 11. Graph. Estimated mean pupil diameter (z-score, conductance) by period.
Time period in the drive significantly affected pupil diameter (χ2(7) = 44.12, p < 0.001). Mean pupil diameter was significantly greater in period 2 than all other periods except 8 (p < 0.05). This suggests that during the first merge of the experimental drive, participants were more alert. In addition, the mean pupil diameter was significantly greater during period 8 than all periods except 2 and 4 (p < 0.05).
The experimental condition did not significantly affect pupil diameter, and the interaction between time period and experimental condition was not significant.
The NASA-TLX assessment indicated that the CACC system with merge assist reduced workload compared with the control condition. However, no differences were found in physiological arousal levels between the experimental groups. However, people can mitigate the tendency toward reduced arousal on long drives by engaging in arousal-stimulating secondary activities. In this experiment, participants were not discouraged from engaging in these activities. While care was also taken to avoid encouraging these activities, participants were told that they could listen to the car radio or do what they normally did while driving.
To explore potential engagement in other arousal-increasing tasks, nondriving activities were recorded during two segments in the drive. The first segment was the 30 s prior to the beginning of the first exit maneuver (exit 4). The following nondriving-related activities were engaged in by at least one participant:
Because most of the observed behaviors were rare, the sum of nondriving activities was analyzed as a function of condition and observation period. Neither experimental condition nor observation period, or their interaction, had a significant effect on the probability of engaging in nondriving activities.
Drivers’ actions during each merge were closely monitored to detect differences in driver behavior both over time and as a result of experimental condition. These behaviors included merge success and position, gap selection, and the distance used to complete the merge. The following analyses are not based on the eight previously defined driving segments but rather on the merges themselves. The beginning of each merge was defined consistently across all participants as the moment when the driver passed a specified point on the onramp (shortly after passing through the signalized intersection); merge endings were defined as the moment when half of the driver’s vehicle was laterally inside the CACC platoon in the main lane of traffic.
As in the real world, a successful merge was defined as one in which the driver avoided colliding with other vehicles. As shown in table 5, several participants experienced a collision in their first merge attempt. The collision rate declined with subsequent merges.
Condition |
Merge 1 |
Merge 2 |
Merge 3 |
Merge 4 |
---|---|---|---|---|
Control |
9 |
2 |
2 |
1 |
CACC without merge assist |
5 |
1 |
1 |
3 |
CACC with merge assist |
0 |
0 |
0 |
0 |
If the drivers in the CACC with merge assist condition did not override the system or lose control of the vehicle, then it was not possible to collide with another vehicle during the merge. As a result, none of the drivers in the CACC with merge assist group collided with another vehicle, and this group was excluded from further analysis. There was no significant difference in collision rates between the control group and CACC without merge assist group. Participants were more likely to experience a crash during the first merge than in the three subsequent merges (Bonferroni correction for multiple comparisons, p < 0.05).
Merge position described the location within the gap between two vehicles in the CACC string into which participants merged. It was defined as the ratio of (1) the distance between the front bumper of the participant’s vehicle and the rear bumper of the vehicle ahead and (2) the distance between the front bumper of the vehicle following the participant and the rear bumper of the vehicle ahead, minus the length of the participant vehicle. Thus, values closer to zero reflected merges closer to the vehicle ahead, a value of 0.5 reflected a perfectly centered participant vehicle, and values approaching 1.0 represented a position close to the trailing vehicle. The algorithm used to control vehicle speed for those drivers in the CACC with merge assist was designed to place participant vehicles equally distant between two vehicles, allowing a simple merge with only lateral adjustment in position.
Figure 12 shows the mean merge gap ratios by condition. On average, participants in the control condition tended to merge in a similar location in the gap to those people driving in the CACC with merge assist group. In contrast, the CACC without merge assist group entered the gap significantly closer to the vehicle in front of the participant vehicle (p < 0.05).
Note: Error bars represent estimated 95-percent confidence limits of the means.
MA = Merge assist.
Figure 12. Graph. Estimated mean merge position by treatment group.
All drivers in the CACC conditions were presented with a continuous string of 1.1-s gaps. Drivers in the control condition were presented with a sequence of nine gaps of varying size that continuously repeated. The size and order of these control condition gaps is shown in table 6. The string into which all drivers merged was moving at a nearly constant 70 mi/h. At that speed, each 0.1-s increment in gap size is equivalent to 10.27 ft.
Sequence Number |
Gap (s) |
---|---|
1 |
0.7 |
2 |
1.1 |
3 |
1.5 |
4 |
0.9 |
5 |
1.4 |
6 |
1.2 |
7 |
0.8 |
8 |
1 |
9 |
1.3 |
The frequency of selection of each gap by control group participants is shown in table 7. As can be seen in table 7, there was a slight tendency for participants to select gaps larger than 1 s (63 percent of the selected gaps were larger than 1 s). However, because the gap sequence was not random, and because the sequence was always triggered by the participant reaching the top of the onramp, this finding could be an artifact (i.e., in each case, the participant chose the gap that was nearest when the participant reached the merge area). As will be seen in the analysis of distance used, there was very little variation in the distance the control group traveled before completing the merges.
The distance required to execute a merge might reflect the ease and/or comfort with which drivers made each merge. A short distance could suggest that the driver easily found an acceptable gap, whereas a longer distance could suggest greater difficulty. Mean distances from the top of the onramp to the completion of the merge, computed by GEEs, are shown in figure 13 as a function of merge number and treatment group.
Note: Error bars represent estimated 95-percent confidence limits of the means.
MA = Merge assist.
Figure 13. Graph. Estimated mean distance used to merge by merge number and experimental condition.
Condition significantly affected the distance used to complete a merge (p = 0.004). On average, both the control group and without merge assist group used more distance to merge than the CACC with merge assist group. This suggests merge assist might increase onramp throughput.
There was a significant interaction between experimental condition and merge number (p = 0.008). This interaction was the result of significant differences between conditions at both the third and fourth merges. In the case of the third merge, the control group used significantly more distance to complete the merge than both the CACC with merge assist group and CACC without merge assist group. At the fourth merge, the control group and the CACC without merge assist group performed similarly, both using significantly more distance to complete the merge than the CACC with merge assist drivers.
Steering entropy is a metric that captures corrective response and has been frequently used to assess driver inattentiveness. Steering entropy was calculated for each subject within each 15-s cruising period (i.e., the periods when drivers were not expected to actively adjust steering to merge). Neither cruising period nor treatment condition yielded significant effects.
One way in which visual attention could be inferred was by examining where drivers were looking. Drivers in the CACC with merge assist group did not need to control speed to successfully merge into the main travel lane. As a result, these drivers might not have felt the need to visually track traffic as closely as the control group and CACC without merge assist groups. This possibility was explored. The proportion of glance time in the direction of the merge area (see figure 14) did not vary significantly among treatment groups.
Both the CACC with and without merge assist groups were required to accept some level of trust in the system. Participants in the CACC with merge assist group were not required to accelerate or brake at any point to successfully complete the merges. Only one participant in this group ever used manual speed controls to override the system. It is not clear, however, whether this participant did not trust the system or simply did not understand how CACC functioned. Throughout much of the drive, that participant manually controlled speed by pressing the accelerator. That participant spun out during the second merge and did not reengage the system during the fourth merge.
Among participants in the CACC without merge assist group, trust was examined only during the cruising periods because these drivers were required to manually control speed when merging. Of the 16 participants in the without merge assist group, 2 engaged the accelerator pedal during a cruise period (1 in period 1 and the other in period 7). However, in both cases, the pedal was used minimally and was possibly the result of resting the foot on the pedal. Thus, it appears that trust in the CACC system was high.
CACC with or without merge assist resulted in about a 50-percent reduction in drivers’ perceived workload. The reduction in workload did not result in a measureable decrease in driver alertness, as assessed by GSR, eyelid opening, or pupil diameter. The merging activity did transiently increase driver arousal when compared with arousal in a period immediately before exiting the string. Thus, the physiological measures used in this study were sensitive to gross changes in demands on driver attention. The lack of an interaction between condition and measurement periods suggests that even with merge assist, attentional demands on drivers were greater during a merge than during uneventful car following.
The lack of a difference in the number of nondriving activities that drivers in the three treatment conditions engaged in suggests that CACC did not relieve drivers of so much workload that they felt compelled to engage in additional activities to manage their level of arousal.
Merge assist, as defined in this study, has not been widely discussed as a driving task to be automated. The elimination of crashes in the absence of indications of changes in visual scanning behavior suggests that this is an area that should be given higher priority for further development. Furthermore, the present finding that those with merge assist required significantly less distance to complete their merges suggests that traffic operations could benefit from merge assist automation.
The rarity of cases in which participants disengaged the CACC system suggests a reasonable level of trust in the system.