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-04-134
Date: December 2005
Enhanced Night Visibility Series, Volume III: Phase II—Study 1: Visual Performance During Nighttime Driving in Clear Weather
PDF Version (859 KB)
PDF files can be viewed with the Acrobat® Reader®
CHAPTER 4—DISCUSSION AND CONCLUSIONS
As mentioned in the Methods section (chapter 2), the headlamp aiming protocol used for this study resulted in a deviation in the maximum intensity location from its typical placement for some headlamp types. Details about this deviation are discussed in ENV Volume XVII, Characterization of Experimental Vision Enhancement Systems. As a result of the headlamp aiming, the presented detection and recognition distances were likely increased for the HLB and HOH configurations and likely decreased for the HHB configuration. The results of this study should be considered in the context and conditions tested. If different halogen headlamps or aiming methods are used, the results might be different.
Detection and recognition distances varied significantly among different VESs during nighttime driving in the clear weather condition. Throughout this discussion, the HLB system will be used as a baseline because of its widespread availability. In this particular study, several systems under- or over-performed the HLB system by as much as 30.5 m (100 ft) (table 17), representing a 19 percent difference. These differences in distance can be translated to gains or losses in reaction time (table 18). Reaction time has been used in the past to evaluate time margins for crash avoidance behavior when encountering obstacles in the driving path.(19) Overall, use of the IR–TIS resulted in significant detection improvements over other systems. Specifically, participants were able to detect objects 24.7 m (81 ft) farther (i.e., a 13 percent increase in distance) with the IR–TIS than with the HLB. On average, the HID configuration provided the lowest detection and recognition distances. When compared to the HLB, the HID headlamps resulted in object detection distances that were 30.2 m (99 ft) closer to the object of interest (i.e., a 16 percent decrease in distance).
While these distances and reaction times provide an indication of the advantages of one system over another, they fail to describe completely any potential safety benefits or concerns based on VES use. With a limited number of assumptions, however, the VES-specific detection distances under clear weather conditions can be compared against various speed-dependent stopping distances. Collision avoidance research dealing with different aspects of visibility suggests that time-to-collision is an important parameter in the enhancement of driving safety.(20) For consistency, time-to-collision is presented as “distance-to-collision” (or stopping distance) for direct comparisons to the detection distances from the current study. Stopping distance is the sum of two components: (1) the distance needed for the braking reaction time (BRT), and (2) braking distance (table 19). Braking distance is the distance that a vehicle travels while slowing to a complete stop.(21) For a vehicle that uniformly decelerates to a stop, the braking distance (dBD) is dependent upon initial velocity (V), gravitational acceleration (g), coefficient of friction (f) between the vehicle tires and the pavement, and the gradient (G) of the road surface, with the gradient measured as a percent of slope. The equation in figure 27 provides the calculation of the braking distance (dBD) under these conditions:
Figure 27. Equation. Braking distance.
The total stopping distance (d) is the sum of the braking distance (dBD) and the distance traveled during the brake reaction time. The results from driver braking performance studies suggest that the 95th percentile BRT to an unexpected object scenario in open road conditions is about 2.5 s. (See references 22, 23, 24, and 25.) For a vehicle traveling at a uniform velocity, the distance traveled during BRT is the product of the reaction time and the velocity. Assuming a straight, level road with a gradient of zero percent (G = 0), the equation for the total stopping distance is as shown in figure 28:
Figure 28. Equation. Total stopping distance for brake reaction time plus braking distance.
The equation in figure 28 may be used with either metric or English units, with distance (d) in meters or feet, velocity (V) in m/s or ft/s, and a value for the acceleration due to gravity (g) of 9.8 m/s2 or 32.2 ft/s2.
The American Association of State Highway and Transportation Officials (AASHTO) provides separate equations for stopping distance with metric and English units, in which the acceleration due to gravity (g) and the coefficient of friction (f) are combined into a deceleration rate, and the velocity (V) is in units of km/h or mi/h, respectively.(22) The equation in figure 28 was used in this report because it does not require conversion factors and allows for a more direct comparison of the effect of varying the coefficient of friction (f).
To calculate total stopping distance, AASHTO suggests using a deceleration rate (a) of 11.2 ft/s2 (3.4 m/s2), resulting in a friction coefficient for wet pavement of 0.35 as seen in the equation in figure 29.(22)
Figure 29. Equation. AASHTO calculation of coefficient of friction for wet pavement.
The coefficient of friction used for these calculations is based on Lindeburg data for dry surface conditions.(26) The data obtained from Lindeburg is comprehensive in terms of type of surface, tire condition, and speed. A mean value of 0.65 was obtained for the coefficient of friction for dry surfaces (across all dry conditions). To accommodate most types of vehicles’ braking capabilities, a conservative approach was taken for the calculations, and 0.60 was used as the coefficient of friction. Using this approach, stopping distances were calculated as shown in table 19.
The calculations represent a simple and ideal condition, but the formula allows for some visualization of the capabilities VES has. Based on these calculations, the average detection distances for each IR–TIS and HLB VES (table 17) provide enough time to react and brake up to speeds of less than approximately 105 km/h (65 mi/h). HID configurations supplemented with UV–A, HOH, and HHB show detection distances that will allow braking for up to 89 km/h (55 mi/h). The only two VESs that seem to be ineffective at more than 89 km/h (55 mi/h) are HID and HLB–LP; however, some caveats apply. First, these distances were obtained while drivers were moving at approximately 40 km/h (25 mi/h), and their ability to detect objects will not necessarily remain the same as speed increases. Second, VESs that are currently close to the stopping distance or that need a larger stopping distance (e.g., HID, HLB–LP) might quickly become more ineffective when conditions worsen (e.g., wet pavement, worn tires, downhill condition). Third, and most important, when detection distances are analyzed in more detail by examining the significant (p < 0.05) VES by Object interaction, different conclusions can be reached. (In table 20 through table 31, an “X” means the stopping distance might be compromised; an asterisk means the same thing, but in an unlikely scenario.) Several VES and object combinations resulted in detection distances that might compromise stopping distances.
The literature review performed as part of the larger ENV project suggested that new headlamps (such as high intensity discharge, configurations supplemented by UV–A headlamps, and infrared thermal imaging systems) could be expected to outperform halogen headlamps (ENV Volume II). These expectations were not completely fulfilled.
As expected, the infrared technology allowed the detection of warm objects (i.e., pedestrians and cyclists) at distances of 201.2 to 292.3 m (660 to 959 ft), an improvement of more than 76.2 m (250 ft) beyond the halogen headlamps for dark-clothed perpendicular pedestrians. This improvement over HLB is consistent with results obtained by Barham et al.(9) Interestingly, the improvement obtained from infrared technology for pedestrians on the side of the road (e.g., a person on the side of the road waiting to cross the street, or static pedestrian) is not as dramatic: only 2.4 to 4 m (8 to 13 ft). It is possible that the size of the HUD does not allow for a sufficiently broad field of view. Participants tended to support this theory during interviews, with comments such as: “I thought that it kind of needed to be a little broader field. I felt like it cut down a little bit on my peripheral (vision),” referring to the image on the IR–TIS.
Jost suggested that HID systems should improve visibility distance by more than 50 percent compared to standard HLB systems.(27) The HID system used for this study did not perform up to this expectation. In fact, detection distances for the individual objects with this HID system were 8.5 to 54.9 m (28 to 180 ft) closer to the object that needed to be detected than were distances obtained using halogen headlamps. It is possible that the HID system tested here differed significantly in terms of cutoff and intensity from the HID systems tested in other investigations.
The characteristics of these systems vary considerably among manufacturers. While unpublished data generated by this investigation (refer to ENV Volume XVII) agree with Jost(27) that HIDs provided more luminous flux than regular halogen headlamps, the problem with the current HID system involves where the luminous flux is directed. The large amount of visible light generated by these systems requires a dramatic cutoff angle to comply with glare standards. While this provides more foreground luminance, less illumination is provided as the distance from the vehicle increases. This foreground luminance might affect driver performance by increasing the driver’s light adaptation, thus decreasing the driver’s capability to detect objects in dark
Mahach et al.(28) and Nitzburg et al.(29) suggest that UV–A could improve visibility distances. This previous research on pedestrian visibility was performed in a static environment (i.e., the car’s transmission was in the “park” position), and the participants were in the passenger side of the vehicle. Between detection and recognition trials, the vehicle moved in increments of 30.5 m (100 ft). A windshield shutter was used to limit the time available for visual search, and a 2-s stimulus exposure time was given for each trial (i.e., each time the vehicle moved 30.5 m (100 ft)). Results suggested improvements in visibility distances by more than 200 percent when UV–A detection distances were compared to halogen headlamp detection distances.
The current results dispute this finding; however, a comparison of similar trials from both studies can provide some information on the reasons for the apparent decrease in performance of UV–A technology in the current study. One reason why UV–A configurations did not result in a 200 percent improvement over HLB in this study might be that the halogen headlamp technology used was dramatically different from the one used in the Mahach et al.(28) study (figure 30). For example, the static pedestrian was a common object in Mahach et al.(28) and the current investigation. Note that static pedestrian detection distances obtained for Mahach’s “HLB-May 97” were 191.4 and 176.2 m (628 and 578 ft) smaller for the HLB and HLB–LP systems, respectively, than the ones tested in this study. While an improvement in detection distance occurred with the UV–A technology used in the current study (compared to the UV–A technology used for the Mahach et al.(28) study), it was not large enough for the UV–A technology to maintain the advantage over the HLB and HID systems used in this study. Another possibility is that given the limited amount of environmental exposure time in the Mahach et al. study (2-s window exposure every 30.5 m (100 ft)), those distances represent less of an absolute threshold, whereas the focus of the current investigation is to identify the exact distance that will produce detection during a dynamic condition.
Figure 30. Bar graph. Comparison of the results obtained for UV–A headlamps with previous research.
Depending on the VES, age was responsible for some of the variability in detection distances; however, this was not the case for recognition distances by VES, where the variability between age groups was not significant. On average, detection distances by VES for the younger drivers ranged from 168.6 to 223.1 m (553 to 732 ft), whereas detection distances for older drivers ranged from 137.5 to 179.8 m (451 to 590 ft). Across VESs, detection distances for the older drivers were consistently smaller than for the other age groups. The range of detection distances for middle-aged drivers was similar to distances for the younger drivers, 157 to 231 m (515 to 758 ft). These differences can be quantified in terms of a VES baseline (HLB, table 32), and in terms of the pair-wise differences between the three groups (table 33).
The IR–TIS resulted in detection distances of more than 30.5 m (100 ft) longer than those obtained with the HLB for the younger and middle-aged groups. The detection distances for older drivers using the IR–TIS were the same as the distances obtained by this group using HLBs and only 24.1 m (79 ft) longer than the distances that this group obtained using HLB–LP. This difference between age groups might be the result of the information processing nature of the HUD task. Although all drivers were equally trained, older people in general take longer to retrieve information, and time-sharing among tasks tends to pose a greater informational demand.(30,31) In addition, HUD users risk cognitive capture, which might occur when there is inefficient switching of attention between the HUD and the external environment.(32,33) Inefficient switching is of paramount importance because it may result in missed external objects and delayed responses. It is possible that older drivers in this experiment were less efficient than those in the other two age groups at switching from the HUD to the task, or vice versa. Moreover, they might not have used the IR–TIS at all. Some drivers demonstrated concern about the time-sharing demand of the HUD during the interviews:
“You felt like you had two things to look at. It’s only a small image in front of you, but yet you have the entire picture on the windshield that you are trying to look at too, so you’re afraid that if you just look at the image you might be missing something; it may not be broad enough to see something that’s really out there, if you just looked at one or the other it would be a little bit different, but when you have the choice of looking at one or the other you feel like something may appear in the picture that you don’t really see...” (Participant #42-middle—aged female)
“...it is kind of down below and you don’t know whether you should try to like drive with it, or look ahead and just kind of glance down there every once in a while. Or you should just look ahead and not use it, and just use if you see something flashing through there, then you look down at it... it was a little confusing to get used to; I mean I definitely think it’s cool, but it was kind of down low and I kind of want to scrunch down to try to look through it.” (Participant #37—younger male)
When the average detection distances for the three groups by VES configuration are compared to the stopping distances needed (table 19) for a highway-type environment (i.e., 105 km/h (65 mi/h)), the maximum stopping distance is close to the detection distance observed for older drivers when HIDs or HLB–LPs were used.
Age caused significant differences on detection and recognition distances depending on the type of object (table 34 and table 35). Older drivers appeared less capable of detecting and recognizing low-contrast objects than their younger counterparts. In fact, the ability of older drivers to detect and recognize pedestrians and cyclists with black clothing was reduced by 13 to 21 percent when compared to the abilities of the younger drivers. This difference in performance is likely the result of the decrease in visual acuity and contrast sensitivity that occurs with age. The decline generally begins slowly after 40, followed by an accelerated decline after 60.(2,3,5) This trend was observed between the various age groups in this investigation. Figure 31 shows participants’ visual acuity, and figure 32 through figure 36 show participants’ percentage of contrast for the left eye (PCL) and right eye (PCR) for test lines A through E, which represent 1.5, 3.0, 6.0, 12.0, and 18.0 cycles per degree (cpd), respectively.
Figure 31. Bar graph. Participants’ visual acuity divided by age group.
Figure 32. Bar graph. Participants’ contrast sensitivity at 1.5 cpd (cycles per degree)
Figure 33. Bar graph. Participants’ contrast sensitivity at 3.0 cpd divided by age group.
Figure 34. Bar graph. Participants’ contrast sensitivity at 6.0 cpd divided by age group.
Figure 35. Bar graph. Participants’ contrast sensitivity at 12.0 cpd divided by age group.
Figure 36. Bar graph. Participants’ contrast sensitivity at 18.0 cpd divided by age group.
Comparisons were made in Study 1 to determine whether VESs that show an increase in detection and recognition distances for pedestrians and cyclists also show the same trend for other objects such as the tire tread and the child’s bicycle. HLB headlamps are used in this comparison as a baseline system. Table 36 and table 39 give the mean detection and recognition distances, respectively, for all VESs and objects. The top three detection and recognition distances for each object are highlighted in table 37, table 38, table 40, and table 41 (1st = green, *; 2nd = blue, **; 3rd = yellow, ***).
While no single VES was superior across all objects, HLB and HLB with UV–A consistently provided the driver with the one of the top (farther away from the object) detection and recognition distances across all objects in clear weather conditions. The best VES for the detection of pedestrians and cyclists with black (low contrast) clothing (i.e., IR–TIS) did not perform as well for objects at ground level (i.e., tire tread and the child’s bicycle). For pedestrians in white clothing, the IR–TIS allowed early detection of pedestrians in front of the vehicle path (i.e., perpendicular pedestrian in white clothing), but not those on the side of the road (i.e., static pedestrian in white clothing and parallel pedestrian in white clothing). This effect is probably the result of the system’s limited field of view (limited size of the screen). The small screen required by the HUD application limits the amount of information that can be displayed while maintaining an acceptable level of resolution.
HLB, by itself or supplemented by UV–A, always placed either first or second across all different objects in both detection and recognition. The additional effect of UV–A was relatively small for pedestrians and nonmotorists with black clothing (i.e., less than 7 percent detection increase over HLB). However, HLB supplemented with the UV–A headlamps represented up to an 11 percent (i.e., 28.7 m (94 ft)) increase in detection distance and a 13 percent (i.e., 29.3 m (96 ft)) increase in recognition distances over HLB for pedestrians and nonmotorists with white clothing. Two of the UV–A supplemented HID configurations performed well for detection and recognition of high contrast objects on the side of the road (i.e., static and parallel pedestrians with white clothing). This result was not surprising because the HID headlamp beam pattern had increased illumination toward the driver’s right side, the same spot in the road where these pedestrians were positioned. While HLB and HLB supplemented by UV–A were always in the first three positions (best) for both detection and recognition, the rankings were modified in some instances. For example, IR–TIS was the best for detecting the black-clothed cyclist and white-clothed perpendicular pedestrian, but it was not the best for recognition of these objects.
Several interesting aspects are found when the detection distance obtained with the IR–TIS for pedestrians with white (high contrast) versus black (low contrast) clothing is compared (table 36). Pedestrians dressed with dark clothing were not detected by the driver as far away as the ones with high contrast clothing. Pedestrians with dark clothing were detected more than 57.9 m (190 ft) closer than the ones with white clothing. This could be because some drivers are trying to find additional visual cues from the windshield (in addition to the IR–TIS image) that cannot be captured beyond 243.8 m (800 ft) due to the low contrast of the object. However, even for low contrast objects, their detection increased by more than 60 percent of the distances at which the objects can be detected with just halogen headlamps.
In summary, during the clear weather condition, no VES is consistently the best in facilitating long detection and recognition distances. In addition, both the HLB and HID baseline headlamps indicated little benefit from the additional UV–A sources, although the aiming protocol used for this study likely increased detection and recognition distances for the mechanically aimed HLB headlamps. The following conclusions can be reached regarding Study 1 of Phase II: