- Email: [email protected]
On the highway measures of driver glance behavior with an example automobile navigation system
ARTICLE IN PRESS
Applied Ergonomics 35 (2004) 215–223
On the highway measures of driver glance behavior with an example automobile navigation system Dean P. Chiang*, Aaron M. Brooks, David H. Weir Dynamic Research, Inc., 355 Van Ness Avenue, Torrance, CA 90501 USA
Abstract An over-the-road study of visual–manual destination entry using an example original equipment GPS-based navigation system was accomplished in traffic on urban streets and motorways. The evaluation used typical drivers, and a vehicle instrumented to record driver eye glances and fixations, driver control inputs, and lateral lane position. The primary task was to drive in a safe manner, in traffic, while maintaining speed and lateral lane position. As a secondary task, the drivers entered successive destinations while driving, using a touchscreen, and at their own pace. They were told there was no need to enter the destination quickly. Results are shown for driver glance behavior, lane keeping performance, and subjective ratings. Overall, the drivers were able to accomplish the destination entry tasks with acceptably short glance durations, acceptable total task times, and with satisfactory subjective ratings for ease of entry. r 2004 Elsevier Ltd. All rights reserved. Keywords: Navigation systems; Driver distraction; Driver glance behavior
1. Introduction Traditionally, drivers have operated devices in the vehicle, such as radios, heating and air conditioning, and other auxiliary controls as secondary tasks; in addition to the primary task of following the prescribed path at the desired speed. Historical accident data and anecdotal evidence suggest that drivers have been able to adapt to these additional attentional workloads without affecting primary task performance unduly. More recently, new in-vehicle devices have been introduced such as cellular telephones, more complex entertainment systems, navigation systems, and other devices aimed at assisting the driver. While these devices are all useful in one or more important ways, they raise the issue of possible driver distraction and whether the secondary workload can become sufficiently high that it begins to affect primary task performance. Of particular interest is the task of navigation system destination entry. Several studies and literature reviews (e.g., Green, 1999) have been accomplished that relate to this topic, and it is under consideration by several technical *Corresponding author. Tel.: +1-310-212-5211; fax: +1-310-2125046. E-mail address: [email protected] (D.P. Chiang). 0003-6870/$ - see front matter r 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.apergo.2004.01.005
committees and organizations (e.g., SAE, 2002; ISO, 2002; AAM, 2002). A study reported by Tijerina, 2000, measured route guidance system destination entry while driving (in their case, on a test track). Recent work accomplished by JARI for JAMA (Hashimoto and Atsumi, 2001) studied destination entry tasks on roads in Japan. While the results of several over-the-road studies involving route following have been reported, to our knowledge, other than the work cited, at the time this project was initiated little had been published on visual–manual destination entry under over-the-road conditions in traffic. As a result, this study observed and measured the use by typical drivers in over-the-road conditions of an example contemporary navigation system with regards to visual–manual destination entry.
2. Method The participant drivers entered destinations using the touchscreen display of a navigation system while driving an instrumented vehicle on city streets and on an urban motorway. The destination entry keystrokes and associated eye movements were videotaped and later reduced to obtain keystroke timing, glance behavior, entry errors, and other data.
ARTICLE IN PRESS D.P. Chiang et al. / Applied Ergonomics 35 (2004) 215–223
216
Face Camera
Destination card placement
Navigation System
Fig. 1. Driver’s View of Instrument Panel.
first. Therefore, ‘‘Address’’ was selected from the first screen and ‘‘Street’’ from the second screen, as shown in Fig. 2. The next step was to enter the street name, as shown in Fig. 3. This involved entering letters from an alphabetical keypad layout until the screen automatically switched to a list of streets. Then, the street was selected from the list. Fig. 4 shows the screen for entering the numerical address. The numbers are arranged in a conventional telephone keyboard configuration. Once the address was completely entered, a ‘‘Done’’ button was pressed. Next, the city was selected from a list of possibilities, as shown in Fig. 5. A maximum of 4 cities was shown on the screen. The final
2.1. Participants Ten drivers participated in the evaluation. Nine of the drivers were novice users of the navigation system and 1 had used it previously. The driver sample included 5 males and 5 females, 5 of whom were engineers. The others had various administrative and clerical jobs. One of the engineers was female, and another was the oldest participant. Their average age was 33, and their average mileage per year was 13,400 (21,500 km).
Enter destination by: Address
Place
Intersection
2.2. Apparatus
Find address by:
3. Procedure for destination entry All destinations were entered using the ‘‘by Address’’, and ‘‘Street’’ method. This meant that the full address of a destination was entered, with the street name entered
City Street
Fig. 2. Destination Entry Sequence.
Enter street name: HA A B C D E F G H I J K L M N O P
Number
Q R S T U V W X
Delete
Y Z &
/
-
Choose a street: HARPER HATHAWAY HAWTHORNE MADRONA MARICOPA
Space
'
List
V
The example vehicle was a 2000 Acura 3.2 RL equipped with the Acura Navigation System. The navigation display was located relatively high in the center console of the instrument panel, as shown in Fig. 1. All drivers were able to reach the display, comfortably, from their normal seating position and posture. For this study, the touchscreen was used as the entry means, and destinations were entered in this manner. Some error-correction required the use of the CANCEL button to the left of the screen. The destination address was displayed on a card fixed to the steering wheel hub. The vehicle instrumentation consisted of video cameras and recording, and driver control and vehicle response sensors. Two small cameras were mounted on the side mirrors and angled down to view the roadway lane markers, and record lateral lane position. A mini camera was mounted on top of the instrument panel, above the navigation display, pointed towards the face of the driver to record eye movements. A fourth camera was mounted between the driver and passenger seats, pointing at the navigation display to view driver actions and destination inputs.
Fig. 3. Example Street Entry.
V
ARTICLE IN PRESS D.P. Chiang et al. / Applied Ergonomics 35 (2004) 215–223
Enter street number: 3268 1 4 7
2 5 8 0
City Streets
3 6 9
Letter Delete Done Freeway
Fig. 4. Example Numerical Address Entry.
V
Choose a City. Same street name appears in more than one city.
V
LONG BEACH LOS ANGELES SAN DIEGO SAN FRANCISCO Fig. 5. Example City Selection.
Calculate route to: 3268 Harper Dr. Long Beach, CA By
OK
217
Direct Route Method Chg. Method Map
Fig. 6. Example Confirmation Entry.
step was to confirm the address entered, as shown in Fig. 6, by pressing an ‘‘OK’’ button. The combination of a street entry, an address entry, a city selection, and a confirmation entry comprised a destination entry. The destinations used for the practice trials and the evaluation trials represented an average level of difficulty for the task. The street names were chosen based on the number of letters that had to be entered. They were selected such that an average of 5 letters was entered to reach the street list, with an effective range of 4–6 letters. The numerical addresses were chosen to be 4 digits in length, which was felt to be typical. Within a given address, a particular digit only appeared once. For the destinations selected for this evaluation, a total of 15 keystrokes was required to enter a destination, if the subject made no errors or mispresses. For the city selection, the location of the destination city was always the top entry on the list. In general, the destination cities and addresses were not on the study driving route, but were in nearby cities and locations in the greater Los Angeles area. The overall study route is shown in Fig. 7. It consisted of city streets and urban
Fig. 7. Study Routes.
freeways in the Torrance, California area. The 2 city streets routes (Vermont and Normandie) were 1.1 miles (1.8 km) long, connected by streets approximately 0.5 miles (0.8 km) long. Both streets were 2 lanes in each direction, and the drivers drove in the left lane at 35 mph (56 km/h). The urban motorway or freeway sections were 6 miles in length, and the drivers looped back to the original entry point. The freeways were 4 lanes in each direction, and drivers drove in the second lane from the right at 65 mph (105 km/h) or with traffic. The routes were driven on weekdays either in the morning (9:30–11:30 am), or in the afternoon (1:30– 3:30 pm). In general, the 2 sessions had similar traffic conditions. The evaluations took place only in good weather and on dry roads.
4. Experimental procedure The drivers began by reviewing the navigation system user’s manual, and receiving operating instructions. They were told that their primary task was to maintain the vehicle speed, maintain a safe following distance, and keep the vehicle in the lane. The destination entry task was described as less important, and the drivers were told that there was no hurry and that they did not have to enter the destinations quickly. As noted above, the destinations were not immediately ahead on the driving study route. If a driver did not feel comfortable or safe at any time during a destination entry sequence, that run was aborted, which rarely occurred. Prior to the over-the-road evaluations, the drivers first did 3 practice trials with the vehicle parked. Most of them were generally familiar with the area. After driving the study route loop one time, drivers were given 2 practice destinations to enter while driving, one on each stretch of road. The practice trial destinations were different from the evaluation trials. The drivers could ask questions during the practice trials. The same practice trials were used for each, and they were always shown in the same order.
ARTICLE IN PRESS D.P. Chiang et al. / Applied Ergonomics 35 (2004) 215–223
218
After the familiarization and practice trials, the drivers did 3 evaluation trials on the city streets, with one trial on each road section. In order to eliminate possible ordering effects, 2 groups were used. Group A entered the evaluation destinations in the order 1–2–3, while Group B entered the destinations in the order 3–2– 1. After completing the 3 trials on the city streets, the driver stopped the vehicle and filled out a subjective rating form. Then they were directed to the freeway. They first drove the freeway to become familiar with the route. On the second loop, they entered the evaluation destinations. After completing the 3 evaluation trials and exiting the freeway, they filled out another subjective rating form. The entire procedure took approximately 2 h for one driver to complete.
5. Results Learning effects in terms of total task times were examined across all destination entries for both groups of drivers. A 2 factor analysis of variance showed that there were no statistical differences between the 2 driver groups, or differences due to the order in which the destinations were presented. 5.1. Glance behavior results Data reduction from the videos for each over-theroad destination entry included the transition times and glance durations, the number of keystrokes, the number of fixations or ‘‘chunks,’’ and the number of keystroke errors. Eye movement data consisted primarily of the individual fixation times on each of several general ‘‘targets’’ within the driver’s field of view, and the percentage of total fixation times on those targets. The following definitions adapted from SAE, 1999 are used with the glance behavior data: *
*
*
*
*
*
*
Target—the forward road scene, the navigation system display, or other. Fixation time—the duration of a fixation at 1 target. It does not include the transition time to or from the target. Glance duration—fixation time plus the transition to the target, the glance time. Total fixation time—the sum of the individual fixation times on 1 target for a destination entry. Total glance time (TGT)—the sum of the individual glance times on 1 target. Total task time (TTT)—the elapsed time between when a driver first looks away from the road until he or she finally looks back after the last fixation for 1 destination entry. Transition—a change in eye fixation location from 1 target to another.
Table 1 Summary of mean total task times (s) Driving route
Entry destination
Total task time
City streets
1 2 3
34 35 34
5.4 5.7 7.6
Freeway
1 2 3
32 37 34
7.7 7.9 10.9
*
Standard deviation
Entry chunk—1 or more entry keystrokes, or other entry actions, made during one fixation on the display. Attention to the navigation display was only considered to be a chunk if some entry or other action was made.
Table 1 summarizes the mean total task times on city streets and on the freeway for each destination over all drivers. The mean total task time over all drivers was about 34 s, and individual averages ranged from 27 to 49 s. There was very little overall difference between entries made on city streets and those on the freeway, differing by 0.2 s. On city streets, the 3 destinations had very similar average task times. The only noticeable difference was a higher standard deviation for the destination 3 entry, and this was due to one driver having an unusually long task time for one case. On the freeway, there was more variation between the destination entry times, with greater differences between mean total task times and wider ranges for the minimum and maximum values. This increased variability was probably due to the greater attentional demand of the primary task in traffic on the freeway. An analysis of variance was performed to check for statistical differences in the Table 1 entry times between the destinations, and there were no significant differences at the 95% confidence level. An overall average of about 17 keystrokes was used to enter the destinations (17.1 on city streets, 17.4 on freeway). This ranged from 15 to 25 with 15 keystrokes being the necessary minimum in each case. All occurrences of more than 15 keystrokes were the result of errors or mispresses. As would be expected, the total task time generally increased when the number of keystrokes increased. One entry chunk involved one fixation on the display, and this consisted of 1 or more keystrokes, errors, or mispresses. Chunks were determined from the eye movements, and the number of keystrokes in a particular chunk was also recorded. Figs. 8 and 9 show the total task times as a function of the number of chunks on city streets and the freeway. Each data point represents one driver and one destination entry. To enter all 15 keystrokes, an overall average
ARTICLE IN PRESS D.P. Chiang et al. / Applied Ergonomics 35 (2004) 215–223
of 13.5 chunks was used (13.1 on city streets, 13.9 on the freeway), with an overall range of 10 to 21 chunks or fixations. When no errors or mispresses occurred during a trial, an overall average of 11.7 chunks was used, ranging from 10 to 13 chunks. So, drivers would sometimes accomplish more than one keystroke in a chunk, and the average number of keystrokes per fixation overall was 1.3.
50
The eye fixations during the destination entry trials were analyzed. There were a total of 799 individual display fixations over all trials and drivers. Table 2 and Figs. 10 and 11 summarize the average total time spent by each driver with eyes fixated on the forward road scene, the navigation display, and other for the city streets and the freeway. ‘‘Other’’ was defined as any fixation time other than the road scene or the navigation display, for example, looking at the side view mirror; plus all of the transition times between fixations. The total fixation time on all 3 parts of the visual field shown here is approximately the same as the total task time.
40 60
30
Mean Fixation Time(sec)
Total Task Time (sec)
60
219
20 10 0 0
5
10 15 Number of Entry Chunks
20
25
50 40
Other
30
Display Fixation
20
Road Fixation
10 0
Fig. 8. Total Task Time Related to Entry Chunks on City Streets.
1
2
3
4
5 6 Subject
7
8
9
10
Fig. 10. Total Fixation Times on City Streets (Mean of 3 Entries). 50 60 Mean Fixation Time (sec)
Total Task Time (sec)
60
40 30 20 10 0 0
5
10 15 Number of Entry Chunks
20
25
Fig. 9. Total Task Time Related to Entry Chunks on Freeway.
50 40
Other Display Fixation Road Fixation
30 20 10 0 1
2
3
4
5 6 Subject
7
8
9
10
Fig. 11. Total Fixation Times on Freeway (Mean of 3 Entries).
Table 2 Summary of fixation percentages on city streets and freeway Fixation percentages Road scene
Navigation display
Other
Subject ID
City streets
Freeway
City streets
Freeway
City streets
Freeway
1 2 3 4 5 6 7 8 9 10
23 22 24 14 22 20 16 27 19 38
24 20 27 17 27 24 26 29 20 37
53 60 54 60 63 53 69 53 55 32
56 59 50 64 55 43 57 50 52 33
24 18 22 26 15 27 15 21 26 30
20 21 23 19 18 33 17 21 27 31
Mean
23
25
55
52
22
23
ARTICLE IN PRESS D.P. Chiang et al. / Applied Ergonomics 35 (2004) 215–223
The difference is that the latter did not include the initial fixation time on the navigation display prior to the first keystroke, nor the remaining fixation time after the last keystroke. In general, the percentage of time the driver spent looking at the road scene during data entry was a little greater on the freeway than on city streets. Since most of the chunks (95%) involved either 1 or 2 keystrokes, Figs. 12 and 13 show distributions of the display fixation times in those cases for city streets and freeways combined. The mean display fixation time was 1.0 s when 1 keystroke was entered in a chunk, and about 1.5 s when 2 keystrokes were entered. Also, 94% of all the display fixations in the cases of 1 or 2
600 Number of Occurrences
220
500 400 300 200 100 0 0
1 2 3 4 Road Scene FixationTime (sec)
5
Fig. 15. Distribution of Road Scene Fixation Times.
100
600
80
500
60
Number of Transitions
Number of Occurrences
120
40 20 0 0
1 2 3 4 Display Fixation Time (sec)
5
200
100
0
20 Number of Occurrences
300
0
Fig. 12. Display Fixation Times for 1 Keystroke Entered.
1
2 3 Transition Time (sec)
4
5
Fig. 16. Distribution of All Transition Times.
15 10 5 0 0
1
2 3 4 Display FixationTime (sec)
5
Fig. 13. Display Fixation Times for 2 Keystrokes Entered.
120 Number of Occurrences
400
100 80 60 40 20 0 0
1
2
3
4
5
Display Fixation Time (sec)
Fig. 14. Display Fixation Times for 1 and 2 Keystrokes Entered.
keystrokes were less than 2.0 s. Combining the results for 1 and 2 keystrokes gives the distribution in Fig. 14. The average display fixation overall was 1.2 s. Fig. 12–14 show fixation times. As defined above, the transition time to the display would need to be added to the fixation time to obtain glance duration. Glance duration is probably the measure of interest when considering possible occlusion methods for evaluating the suitability of navigation displays, such as have been discussed in connection with the Draft SAE Recommended Practice, SAE, 2002. Fig. 15 shows the distribution of the road scene fixation times across all of the drivers for city streets and freeways combined. During the destination entries, the road was fixated for an average of 0.47 s each time, with 95% of the roadway fixations being less than 1.2 s. As noted above, if there were a traffic situation or other primary task event that required an unusually long roadway fixation, then the entry task could be aborted by the driver, but this rarely occurred. Fig. 16 shows the distribution of transition times across all drivers for all transitions (among road scene, display, and other). They averaged 0.15 s, with 95% of the transitions being less than 0.25 s.
ARTICLE IN PRESS D.P. Chiang et al. / Applied Ergonomics 35 (2004) 215–223
221
Table 3 Mean values for glance behavior and task times Driver ID
No. of fixations
Glance duration (s)
Total fixation time (s)
Total glance time (s)
Total task time (s)
1 2 3 4 5 6 7 8 9 10
13.2 11.7 14.2 11.3 13.5 13.4 12.0 13.7 14.5 17.0
1.33 1.46 1.43 1.68 1.28 1.28 1.76 1.25 1.26 0.73
15.2 15.4 17.7 17.1 16.0 14.0 19.5 14.2 15.7 9.2
17.5 17.0 20.2 19.0 17.3 17.1 21.2 15.7 18.2 12.4
31.4 29.2 35.7 30.5 29.8 30.9 35.5 35.3 35.2 47.8
Mean
13.5
1.32
15.4
17.6
34.1
Table 4 Percentile values of glance behavior and task times Mean no. of fixations
City streets Freeway Combined
Total glance time (s)
Total task time (s)
80th
85th
80th
85th
80th
85th
14.0 17.0 15.0
14.7 17.8 16.0
21.1 19.3 20.5
21.3 20.8 21.2
37.9 42.9 40.2
42.0 45.9 43.6
Mean values for glance behavior and total task time are shown in Table 3 for each driver. The values shown for each driver are the averages for the combined data from the city streets and freeways, since these values were not significantly different within a driver. The first 2 columns show the average number of fixations or chunks, and the average duration of each glance. The average total fixation time and the average total glance time (TGT) are shown next, the difference being that the latter include the transitions to the navigation display. The next column is the average total task time while driving over-the-road, and this includes glances to the road ahead, to the navigation display, to other, and all transition times. The total glance time is analogous to the total shutter open time (TSOT) that would be measured if the destination entry task were being accomplished by itself with an occlusion procedure. Table 3 shows some interesting variations for some drivers. For example, driver 7 had 1 very long glance (7.5 s) in one city street trial, which biased her results accordingly. Driver 10 was a non-native English speaker who did not have a large vocabulary of English street names, so he referred more frequently to the cue card to check the spelling, which affected his total task times. He also adopted an entry strategy which consisted of relatively short glances to the display and only one keystroke per glance. The effect of computational delays in the device is included in Table 3. There were typically 5 screen
changes required in 1 trial, each lasting about a second for a total of about 5 s, computationally. While these delays were present in the over-the-road trials, they had only a small effect on the total glance time, because some of the delay occurred when the driver was transitioned away from the navigation display. While average or mean values for glance behavior are shown in Table 3, higher percentile values have been suggested for use in possible design guidelines or as behavioral criteria. Those corresponding 80 and 85 percentile values for these data are shown in Table 4. These are the values below which 80% and 85% percent of the data lie, respectively, and they are, of course, larger than the mean values. For the destination entry tasks used, the percentile values of total glance time are seen to be approximately 20 s. Overall, the results suggest that the total number of glances or chunks is not important, and values greater than 10 are acceptable and routinely used. Of greater importance than the total number of glances is whether the task is chunkable, and can be accomplished in separate incremental steps, with glance durations typically less than 2 s. Similarly, the data suggest that acceptable total task times can be in the 30–40 s range, or greater. Other recent over-the-road results suggesting the desirability of shorter task time limits (e.g., Hashimoto and Atsumi, 2001) may reflect a feeling by their drivers that the secondary entry task had to be completed quickly, as opposed to the results reported
ARTICLE IN PRESS D.P. Chiang et al. / Applied Ergonomics 35 (2004) 215–223
222
Table 5 Lane exceedances associated with destination entry trials Direction
City streets
Freeway
Total
Left Right Total
2 1 3
0 1 1
2 2 4
herein where the driver looked away from the primary task only when they felt comfortable and safe to do so. Finally, the observed total glance times of about 20 s for the example destination entries used in the present study are consistent with the suggested upper value in AAM, 2002. 5.2. Driving performance results During the driving needed to accomplish the experiments, 5 lane exceedances occurred as recorded by the video cameras. A lane exceedance was defined as the outer edge of the tire going beyond the lane edge marker stripe, and typically this involves the tire going only slightly beyond the stripe. Four of the exceedances were associated with a time interval within which a driver was making an experimental run, and these are summarized in Table 5. Two of these 4 lane exceedances occurred on city streets with the subject crossing the left lane-marker. One additional lane exceedance occurred during the total of 3.1 h of driving that were needed to accomplish the experiments, giving a total of 5 exceedances in 3.1 h or approximately 1.6 exceedances/h. To put this observed exceedance rate in perspective, a separate baseline study was conducted to observe and tabulate lane exceedances of the general population of vehicles in typical traffic in over-the-road operations. Video cameras were fitted to an observing vehicle to record the lateral lane position of other vehicles in the lane ahead and the 2 adjacent lanes. The same city street and urban freeway network noted in Fig. 7 was used, with medium traffic conditions (traffic on the freeway moving steadily at speeds of 60–65 mph). The lanes were typically 12 feet wide. An observer in the passenger seat helped to note exceedances by the observed vehicles, and kept a run log. Later, the videos were reviewed frameby-frame to obtain the actual exceedance data. Only passenger cars, small (pickup) trucks and vans, and SUVs were included. A total of 48 other vehicle exceedances was observed in about 20 h of observation. Typically, 1 to 3 other vehicles were being observed at any one time, and video analysis showed that this resulted in the observation of about 30 h of vehicle operation divided roughly into 80% freeway and 20% city streets. The observed exceedances were similarly divided. Dividing 48 by 30 gives about 1.6 exceedances/
vehicle/h as an observed baseline. In most cases it was not possible to determine what, if anything, caused the exceedance of the other vehicle. It could have been an in-vehicle device distraction; a personal reason; driver fatigue, inattention or impairment; or simply other random factors in the operation of the driver/vehicle/ roadway system. Exceedances were somewhat more frequent on curved sections of the freeways. Overall, this baseline rate and the rate observed in the destination entry trials were not particularly different for practical purposes. Vehicle speeds were recorded during the destination entry trials and familiarization runs. The drivers were instructed to drive at a comfortable, constant speed without exceeding the speed limit. On city streets, the average vehicle speed was 37 mph (59 km/h), slightly above the speed limit of 35 mph (56 km/h), with an overall speed range of about 74 mph (76 km/h). The average familiarization (no entry) vehicle speeds on city streets were similar. The average vehicle speed of 57 mph (91 km/h) on the freeway was below the speed limit of 65 mph (104 km/h), having a range of about 75 mph (78 km/h). The familiarization (no entry) vehicle speed on the freeway was higher than during the destination entry trials, by an average of about 3 mph (5 km/h), but the average familiarization speed was also under the speed limit. The extent to which freeway traffic may have influenced the average speeds was not analyzed. 5.3. Subjective rating results Each driver completed several rating forms after the 3 trials on the city streets, and another set after the 3 trials on the freeway. As a typical example, Fig. 17 shows the ratings for Ease of Street Name Entry (alphabetic keypad). Overall, this was rated between Easy and Moderately Easy for both city streets and the freeway.
Effortless
Easy Mode rate
Difficult Impossible City Streets Fig. 17. Ease of Street Name Entry.
Freeway
ARTICLE IN PRESS D.P. Chiang et al. / Applied Ergonomics 35 (2004) 215–223
223
in such areas as voice activation may further simplify the driver’s task or provide for the addition of other invehicle devices and information features, this study suggests that more conventional means of display and visual–manual data entry still represent an acceptable and effective means of interaction.
Effortless Easy Moderate Difficult
References
Impossible City Streets
Freeway
Fig. 18. Ease of Address Entry.
Fig. 18 shows the ratings for Ease of Address Entry (number keypad), which was rated Easy for both road types. Overall, the address was considered easier to input than the street name.
6. Discussion Overall, the results of this study show that drivers are able to make visual–manual destination entries with a contemporary navigation system while operating in actual traffic conditions; with acceptable levels of secondary task loading, path performance, and subjective lack of difficulty. The glance durations and eyes off road time behavior used by the drivers were within previously suggested rules of thumb for safe operation (e.g., Green, 1999; Tijerina, 1999, 2000; Farber, 2000; Greenberg; 2000; Zwahlen, 1988), and total entry times of 30–40 s were acceptable to the drivers when there was no need to enter the destination quickly. While future developments
AAM., 2002. Principles on Human Machine Interface (HMI) for InVehicle Information and Communication Systems (draft). Alliance of Automobile Manufacturers, Detroit. Farber, E., et al., 2000. Surrogate measures of visual demand while driving. Proceedings of the IEA 2000/HFES 2000 Congress, San Diego. Green, P., 1999. Visual and task demands of driver information systems, UMTRI 98-16. University of Michigan. Greenberg, J., 2000. Visual demand while driving, selected results from a Ford Motor Co. driving simulator study. Presented to the SAE ITS Safety and Human Factors Committee. Hashimoto, K., Atsumi, B., 2001. Study of occlusion techniques for making the static evaluation method of visual distraction. Presented at the WG8 Occlusion Workshop, Torino. ISO., 2002. Assessment of visual distraction, preliminary work item, ISO TC22/SC13/WG8, June 2002. SAE.m, 1999. Definition and experimental measures related to the specification of driver visual behavior using video based techniques. SAE J2396, Warrendale, PA. SAE., 2002. Navigation and route guidance function accessibility while driving (draft). Recommended Practice, SAE J2364, Warrendale, PA. Tijerina, L., 1999. Issues in the evaluation of driver distraction associated with in-vehicle information and telecommunication systems. SAE TOPTEC, San Diego, June 1999. Tijerina, L., et al., 2000. Driver distraction with wireless telecommunications and route guidance systems. DOT HS809-069. Zwahlen, H.T., et al., 1988. Safety Aspects of CRT Touch Panel Controls in Automobiles, Vision in Vehicles-II. Elsevier Science Publishers BV, Amsterdam.
Applied Ergonomics 35 (2004) 215–223
On the highway measures of driver glance behavior with an example automobile navigation system Dean P. Chiang*, Aaron M. Brooks, David H. Weir Dynamic Research, Inc., 355 Van Ness Avenue, Torrance, CA 90501 USA
Abstract An over-the-road study of visual–manual destination entry using an example original equipment GPS-based navigation system was accomplished in traffic on urban streets and motorways. The evaluation used typical drivers, and a vehicle instrumented to record driver eye glances and fixations, driver control inputs, and lateral lane position. The primary task was to drive in a safe manner, in traffic, while maintaining speed and lateral lane position. As a secondary task, the drivers entered successive destinations while driving, using a touchscreen, and at their own pace. They were told there was no need to enter the destination quickly. Results are shown for driver glance behavior, lane keeping performance, and subjective ratings. Overall, the drivers were able to accomplish the destination entry tasks with acceptably short glance durations, acceptable total task times, and with satisfactory subjective ratings for ease of entry. r 2004 Elsevier Ltd. All rights reserved. Keywords: Navigation systems; Driver distraction; Driver glance behavior
1. Introduction Traditionally, drivers have operated devices in the vehicle, such as radios, heating and air conditioning, and other auxiliary controls as secondary tasks; in addition to the primary task of following the prescribed path at the desired speed. Historical accident data and anecdotal evidence suggest that drivers have been able to adapt to these additional attentional workloads without affecting primary task performance unduly. More recently, new in-vehicle devices have been introduced such as cellular telephones, more complex entertainment systems, navigation systems, and other devices aimed at assisting the driver. While these devices are all useful in one or more important ways, they raise the issue of possible driver distraction and whether the secondary workload can become sufficiently high that it begins to affect primary task performance. Of particular interest is the task of navigation system destination entry. Several studies and literature reviews (e.g., Green, 1999) have been accomplished that relate to this topic, and it is under consideration by several technical *Corresponding author. Tel.: +1-310-212-5211; fax: +1-310-2125046. E-mail address: [email protected] (D.P. Chiang). 0003-6870/$ - see front matter r 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.apergo.2004.01.005
committees and organizations (e.g., SAE, 2002; ISO, 2002; AAM, 2002). A study reported by Tijerina, 2000, measured route guidance system destination entry while driving (in their case, on a test track). Recent work accomplished by JARI for JAMA (Hashimoto and Atsumi, 2001) studied destination entry tasks on roads in Japan. While the results of several over-the-road studies involving route following have been reported, to our knowledge, other than the work cited, at the time this project was initiated little had been published on visual–manual destination entry under over-the-road conditions in traffic. As a result, this study observed and measured the use by typical drivers in over-the-road conditions of an example contemporary navigation system with regards to visual–manual destination entry.
2. Method The participant drivers entered destinations using the touchscreen display of a navigation system while driving an instrumented vehicle on city streets and on an urban motorway. The destination entry keystrokes and associated eye movements were videotaped and later reduced to obtain keystroke timing, glance behavior, entry errors, and other data.
ARTICLE IN PRESS D.P. Chiang et al. / Applied Ergonomics 35 (2004) 215–223
216
Face Camera
Destination card placement
Navigation System
Fig. 1. Driver’s View of Instrument Panel.
first. Therefore, ‘‘Address’’ was selected from the first screen and ‘‘Street’’ from the second screen, as shown in Fig. 2. The next step was to enter the street name, as shown in Fig. 3. This involved entering letters from an alphabetical keypad layout until the screen automatically switched to a list of streets. Then, the street was selected from the list. Fig. 4 shows the screen for entering the numerical address. The numbers are arranged in a conventional telephone keyboard configuration. Once the address was completely entered, a ‘‘Done’’ button was pressed. Next, the city was selected from a list of possibilities, as shown in Fig. 5. A maximum of 4 cities was shown on the screen. The final
2.1. Participants Ten drivers participated in the evaluation. Nine of the drivers were novice users of the navigation system and 1 had used it previously. The driver sample included 5 males and 5 females, 5 of whom were engineers. The others had various administrative and clerical jobs. One of the engineers was female, and another was the oldest participant. Their average age was 33, and their average mileage per year was 13,400 (21,500 km).
Enter destination by: Address
Place
Intersection
2.2. Apparatus
Find address by:
3. Procedure for destination entry All destinations were entered using the ‘‘by Address’’, and ‘‘Street’’ method. This meant that the full address of a destination was entered, with the street name entered
City Street
Fig. 2. Destination Entry Sequence.
Enter street name: HA A B C D E F G H I J K L M N O P
Number
Q R S T U V W X
Delete
Y Z &
/
-
Choose a street: HARPER HATHAWAY HAWTHORNE MADRONA MARICOPA
Space
'
List
V
The example vehicle was a 2000 Acura 3.2 RL equipped with the Acura Navigation System. The navigation display was located relatively high in the center console of the instrument panel, as shown in Fig. 1. All drivers were able to reach the display, comfortably, from their normal seating position and posture. For this study, the touchscreen was used as the entry means, and destinations were entered in this manner. Some error-correction required the use of the CANCEL button to the left of the screen. The destination address was displayed on a card fixed to the steering wheel hub. The vehicle instrumentation consisted of video cameras and recording, and driver control and vehicle response sensors. Two small cameras were mounted on the side mirrors and angled down to view the roadway lane markers, and record lateral lane position. A mini camera was mounted on top of the instrument panel, above the navigation display, pointed towards the face of the driver to record eye movements. A fourth camera was mounted between the driver and passenger seats, pointing at the navigation display to view driver actions and destination inputs.
Fig. 3. Example Street Entry.
V
ARTICLE IN PRESS D.P. Chiang et al. / Applied Ergonomics 35 (2004) 215–223
Enter street number: 3268 1 4 7
2 5 8 0
City Streets
3 6 9
Letter Delete Done Freeway
Fig. 4. Example Numerical Address Entry.
V
Choose a City. Same street name appears in more than one city.
V
LONG BEACH LOS ANGELES SAN DIEGO SAN FRANCISCO Fig. 5. Example City Selection.
Calculate route to: 3268 Harper Dr. Long Beach, CA By
OK
217
Direct Route Method Chg. Method Map
Fig. 6. Example Confirmation Entry.
step was to confirm the address entered, as shown in Fig. 6, by pressing an ‘‘OK’’ button. The combination of a street entry, an address entry, a city selection, and a confirmation entry comprised a destination entry. The destinations used for the practice trials and the evaluation trials represented an average level of difficulty for the task. The street names were chosen based on the number of letters that had to be entered. They were selected such that an average of 5 letters was entered to reach the street list, with an effective range of 4–6 letters. The numerical addresses were chosen to be 4 digits in length, which was felt to be typical. Within a given address, a particular digit only appeared once. For the destinations selected for this evaluation, a total of 15 keystrokes was required to enter a destination, if the subject made no errors or mispresses. For the city selection, the location of the destination city was always the top entry on the list. In general, the destination cities and addresses were not on the study driving route, but were in nearby cities and locations in the greater Los Angeles area. The overall study route is shown in Fig. 7. It consisted of city streets and urban
Fig. 7. Study Routes.
freeways in the Torrance, California area. The 2 city streets routes (Vermont and Normandie) were 1.1 miles (1.8 km) long, connected by streets approximately 0.5 miles (0.8 km) long. Both streets were 2 lanes in each direction, and the drivers drove in the left lane at 35 mph (56 km/h). The urban motorway or freeway sections were 6 miles in length, and the drivers looped back to the original entry point. The freeways were 4 lanes in each direction, and drivers drove in the second lane from the right at 65 mph (105 km/h) or with traffic. The routes were driven on weekdays either in the morning (9:30–11:30 am), or in the afternoon (1:30– 3:30 pm). In general, the 2 sessions had similar traffic conditions. The evaluations took place only in good weather and on dry roads.
4. Experimental procedure The drivers began by reviewing the navigation system user’s manual, and receiving operating instructions. They were told that their primary task was to maintain the vehicle speed, maintain a safe following distance, and keep the vehicle in the lane. The destination entry task was described as less important, and the drivers were told that there was no hurry and that they did not have to enter the destinations quickly. As noted above, the destinations were not immediately ahead on the driving study route. If a driver did not feel comfortable or safe at any time during a destination entry sequence, that run was aborted, which rarely occurred. Prior to the over-the-road evaluations, the drivers first did 3 practice trials with the vehicle parked. Most of them were generally familiar with the area. After driving the study route loop one time, drivers were given 2 practice destinations to enter while driving, one on each stretch of road. The practice trial destinations were different from the evaluation trials. The drivers could ask questions during the practice trials. The same practice trials were used for each, and they were always shown in the same order.
ARTICLE IN PRESS D.P. Chiang et al. / Applied Ergonomics 35 (2004) 215–223
218
After the familiarization and practice trials, the drivers did 3 evaluation trials on the city streets, with one trial on each road section. In order to eliminate possible ordering effects, 2 groups were used. Group A entered the evaluation destinations in the order 1–2–3, while Group B entered the destinations in the order 3–2– 1. After completing the 3 trials on the city streets, the driver stopped the vehicle and filled out a subjective rating form. Then they were directed to the freeway. They first drove the freeway to become familiar with the route. On the second loop, they entered the evaluation destinations. After completing the 3 evaluation trials and exiting the freeway, they filled out another subjective rating form. The entire procedure took approximately 2 h for one driver to complete.
5. Results Learning effects in terms of total task times were examined across all destination entries for both groups of drivers. A 2 factor analysis of variance showed that there were no statistical differences between the 2 driver groups, or differences due to the order in which the destinations were presented. 5.1. Glance behavior results Data reduction from the videos for each over-theroad destination entry included the transition times and glance durations, the number of keystrokes, the number of fixations or ‘‘chunks,’’ and the number of keystroke errors. Eye movement data consisted primarily of the individual fixation times on each of several general ‘‘targets’’ within the driver’s field of view, and the percentage of total fixation times on those targets. The following definitions adapted from SAE, 1999 are used with the glance behavior data: *
*
*
*
*
*
*
Target—the forward road scene, the navigation system display, or other. Fixation time—the duration of a fixation at 1 target. It does not include the transition time to or from the target. Glance duration—fixation time plus the transition to the target, the glance time. Total fixation time—the sum of the individual fixation times on 1 target for a destination entry. Total glance time (TGT)—the sum of the individual glance times on 1 target. Total task time (TTT)—the elapsed time between when a driver first looks away from the road until he or she finally looks back after the last fixation for 1 destination entry. Transition—a change in eye fixation location from 1 target to another.
Table 1 Summary of mean total task times (s) Driving route
Entry destination
Total task time
City streets
1 2 3
34 35 34
5.4 5.7 7.6
Freeway
1 2 3
32 37 34
7.7 7.9 10.9
*
Standard deviation
Entry chunk—1 or more entry keystrokes, or other entry actions, made during one fixation on the display. Attention to the navigation display was only considered to be a chunk if some entry or other action was made.
Table 1 summarizes the mean total task times on city streets and on the freeway for each destination over all drivers. The mean total task time over all drivers was about 34 s, and individual averages ranged from 27 to 49 s. There was very little overall difference between entries made on city streets and those on the freeway, differing by 0.2 s. On city streets, the 3 destinations had very similar average task times. The only noticeable difference was a higher standard deviation for the destination 3 entry, and this was due to one driver having an unusually long task time for one case. On the freeway, there was more variation between the destination entry times, with greater differences between mean total task times and wider ranges for the minimum and maximum values. This increased variability was probably due to the greater attentional demand of the primary task in traffic on the freeway. An analysis of variance was performed to check for statistical differences in the Table 1 entry times between the destinations, and there were no significant differences at the 95% confidence level. An overall average of about 17 keystrokes was used to enter the destinations (17.1 on city streets, 17.4 on freeway). This ranged from 15 to 25 with 15 keystrokes being the necessary minimum in each case. All occurrences of more than 15 keystrokes were the result of errors or mispresses. As would be expected, the total task time generally increased when the number of keystrokes increased. One entry chunk involved one fixation on the display, and this consisted of 1 or more keystrokes, errors, or mispresses. Chunks were determined from the eye movements, and the number of keystrokes in a particular chunk was also recorded. Figs. 8 and 9 show the total task times as a function of the number of chunks on city streets and the freeway. Each data point represents one driver and one destination entry. To enter all 15 keystrokes, an overall average
ARTICLE IN PRESS D.P. Chiang et al. / Applied Ergonomics 35 (2004) 215–223
of 13.5 chunks was used (13.1 on city streets, 13.9 on the freeway), with an overall range of 10 to 21 chunks or fixations. When no errors or mispresses occurred during a trial, an overall average of 11.7 chunks was used, ranging from 10 to 13 chunks. So, drivers would sometimes accomplish more than one keystroke in a chunk, and the average number of keystrokes per fixation overall was 1.3.
50
The eye fixations during the destination entry trials were analyzed. There were a total of 799 individual display fixations over all trials and drivers. Table 2 and Figs. 10 and 11 summarize the average total time spent by each driver with eyes fixated on the forward road scene, the navigation display, and other for the city streets and the freeway. ‘‘Other’’ was defined as any fixation time other than the road scene or the navigation display, for example, looking at the side view mirror; plus all of the transition times between fixations. The total fixation time on all 3 parts of the visual field shown here is approximately the same as the total task time.
40 60
30
Mean Fixation Time(sec)
Total Task Time (sec)
60
219
20 10 0 0
5
10 15 Number of Entry Chunks
20
25
50 40
Other
30
Display Fixation
20
Road Fixation
10 0
Fig. 8. Total Task Time Related to Entry Chunks on City Streets.
1
2
3
4
5 6 Subject
7
8
9
10
Fig. 10. Total Fixation Times on City Streets (Mean of 3 Entries). 50 60 Mean Fixation Time (sec)
Total Task Time (sec)
60
40 30 20 10 0 0
5
10 15 Number of Entry Chunks
20
25
Fig. 9. Total Task Time Related to Entry Chunks on Freeway.
50 40
Other Display Fixation Road Fixation
30 20 10 0 1
2
3
4
5 6 Subject
7
8
9
10
Fig. 11. Total Fixation Times on Freeway (Mean of 3 Entries).
Table 2 Summary of fixation percentages on city streets and freeway Fixation percentages Road scene
Navigation display
Other
Subject ID
City streets
Freeway
City streets
Freeway
City streets
Freeway
1 2 3 4 5 6 7 8 9 10
23 22 24 14 22 20 16 27 19 38
24 20 27 17 27 24 26 29 20 37
53 60 54 60 63 53 69 53 55 32
56 59 50 64 55 43 57 50 52 33
24 18 22 26 15 27 15 21 26 30
20 21 23 19 18 33 17 21 27 31
Mean
23
25
55
52
22
23
ARTICLE IN PRESS D.P. Chiang et al. / Applied Ergonomics 35 (2004) 215–223
The difference is that the latter did not include the initial fixation time on the navigation display prior to the first keystroke, nor the remaining fixation time after the last keystroke. In general, the percentage of time the driver spent looking at the road scene during data entry was a little greater on the freeway than on city streets. Since most of the chunks (95%) involved either 1 or 2 keystrokes, Figs. 12 and 13 show distributions of the display fixation times in those cases for city streets and freeways combined. The mean display fixation time was 1.0 s when 1 keystroke was entered in a chunk, and about 1.5 s when 2 keystrokes were entered. Also, 94% of all the display fixations in the cases of 1 or 2
600 Number of Occurrences
220
500 400 300 200 100 0 0
1 2 3 4 Road Scene FixationTime (sec)
5
Fig. 15. Distribution of Road Scene Fixation Times.
100
600
80
500
60
Number of Transitions
Number of Occurrences
120
40 20 0 0
1 2 3 4 Display Fixation Time (sec)
5
200
100
0
20 Number of Occurrences
300
0
Fig. 12. Display Fixation Times for 1 Keystroke Entered.
1
2 3 Transition Time (sec)
4
5
Fig. 16. Distribution of All Transition Times.
15 10 5 0 0
1
2 3 4 Display FixationTime (sec)
5
Fig. 13. Display Fixation Times for 2 Keystrokes Entered.
120 Number of Occurrences
400
100 80 60 40 20 0 0
1
2
3
4
5
Display Fixation Time (sec)
Fig. 14. Display Fixation Times for 1 and 2 Keystrokes Entered.
keystrokes were less than 2.0 s. Combining the results for 1 and 2 keystrokes gives the distribution in Fig. 14. The average display fixation overall was 1.2 s. Fig. 12–14 show fixation times. As defined above, the transition time to the display would need to be added to the fixation time to obtain glance duration. Glance duration is probably the measure of interest when considering possible occlusion methods for evaluating the suitability of navigation displays, such as have been discussed in connection with the Draft SAE Recommended Practice, SAE, 2002. Fig. 15 shows the distribution of the road scene fixation times across all of the drivers for city streets and freeways combined. During the destination entries, the road was fixated for an average of 0.47 s each time, with 95% of the roadway fixations being less than 1.2 s. As noted above, if there were a traffic situation or other primary task event that required an unusually long roadway fixation, then the entry task could be aborted by the driver, but this rarely occurred. Fig. 16 shows the distribution of transition times across all drivers for all transitions (among road scene, display, and other). They averaged 0.15 s, with 95% of the transitions being less than 0.25 s.
ARTICLE IN PRESS D.P. Chiang et al. / Applied Ergonomics 35 (2004) 215–223
221
Table 3 Mean values for glance behavior and task times Driver ID
No. of fixations
Glance duration (s)
Total fixation time (s)
Total glance time (s)
Total task time (s)
1 2 3 4 5 6 7 8 9 10
13.2 11.7 14.2 11.3 13.5 13.4 12.0 13.7 14.5 17.0
1.33 1.46 1.43 1.68 1.28 1.28 1.76 1.25 1.26 0.73
15.2 15.4 17.7 17.1 16.0 14.0 19.5 14.2 15.7 9.2
17.5 17.0 20.2 19.0 17.3 17.1 21.2 15.7 18.2 12.4
31.4 29.2 35.7 30.5 29.8 30.9 35.5 35.3 35.2 47.8
Mean
13.5
1.32
15.4
17.6
34.1
Table 4 Percentile values of glance behavior and task times Mean no. of fixations
City streets Freeway Combined
Total glance time (s)
Total task time (s)
80th
85th
80th
85th
80th
85th
14.0 17.0 15.0
14.7 17.8 16.0
21.1 19.3 20.5
21.3 20.8 21.2
37.9 42.9 40.2
42.0 45.9 43.6
Mean values for glance behavior and total task time are shown in Table 3 for each driver. The values shown for each driver are the averages for the combined data from the city streets and freeways, since these values were not significantly different within a driver. The first 2 columns show the average number of fixations or chunks, and the average duration of each glance. The average total fixation time and the average total glance time (TGT) are shown next, the difference being that the latter include the transitions to the navigation display. The next column is the average total task time while driving over-the-road, and this includes glances to the road ahead, to the navigation display, to other, and all transition times. The total glance time is analogous to the total shutter open time (TSOT) that would be measured if the destination entry task were being accomplished by itself with an occlusion procedure. Table 3 shows some interesting variations for some drivers. For example, driver 7 had 1 very long glance (7.5 s) in one city street trial, which biased her results accordingly. Driver 10 was a non-native English speaker who did not have a large vocabulary of English street names, so he referred more frequently to the cue card to check the spelling, which affected his total task times. He also adopted an entry strategy which consisted of relatively short glances to the display and only one keystroke per glance. The effect of computational delays in the device is included in Table 3. There were typically 5 screen
changes required in 1 trial, each lasting about a second for a total of about 5 s, computationally. While these delays were present in the over-the-road trials, they had only a small effect on the total glance time, because some of the delay occurred when the driver was transitioned away from the navigation display. While average or mean values for glance behavior are shown in Table 3, higher percentile values have been suggested for use in possible design guidelines or as behavioral criteria. Those corresponding 80 and 85 percentile values for these data are shown in Table 4. These are the values below which 80% and 85% percent of the data lie, respectively, and they are, of course, larger than the mean values. For the destination entry tasks used, the percentile values of total glance time are seen to be approximately 20 s. Overall, the results suggest that the total number of glances or chunks is not important, and values greater than 10 are acceptable and routinely used. Of greater importance than the total number of glances is whether the task is chunkable, and can be accomplished in separate incremental steps, with glance durations typically less than 2 s. Similarly, the data suggest that acceptable total task times can be in the 30–40 s range, or greater. Other recent over-the-road results suggesting the desirability of shorter task time limits (e.g., Hashimoto and Atsumi, 2001) may reflect a feeling by their drivers that the secondary entry task had to be completed quickly, as opposed to the results reported
ARTICLE IN PRESS D.P. Chiang et al. / Applied Ergonomics 35 (2004) 215–223
222
Table 5 Lane exceedances associated with destination entry trials Direction
City streets
Freeway
Total
Left Right Total
2 1 3
0 1 1
2 2 4
herein where the driver looked away from the primary task only when they felt comfortable and safe to do so. Finally, the observed total glance times of about 20 s for the example destination entries used in the present study are consistent with the suggested upper value in AAM, 2002. 5.2. Driving performance results During the driving needed to accomplish the experiments, 5 lane exceedances occurred as recorded by the video cameras. A lane exceedance was defined as the outer edge of the tire going beyond the lane edge marker stripe, and typically this involves the tire going only slightly beyond the stripe. Four of the exceedances were associated with a time interval within which a driver was making an experimental run, and these are summarized in Table 5. Two of these 4 lane exceedances occurred on city streets with the subject crossing the left lane-marker. One additional lane exceedance occurred during the total of 3.1 h of driving that were needed to accomplish the experiments, giving a total of 5 exceedances in 3.1 h or approximately 1.6 exceedances/h. To put this observed exceedance rate in perspective, a separate baseline study was conducted to observe and tabulate lane exceedances of the general population of vehicles in typical traffic in over-the-road operations. Video cameras were fitted to an observing vehicle to record the lateral lane position of other vehicles in the lane ahead and the 2 adjacent lanes. The same city street and urban freeway network noted in Fig. 7 was used, with medium traffic conditions (traffic on the freeway moving steadily at speeds of 60–65 mph). The lanes were typically 12 feet wide. An observer in the passenger seat helped to note exceedances by the observed vehicles, and kept a run log. Later, the videos were reviewed frameby-frame to obtain the actual exceedance data. Only passenger cars, small (pickup) trucks and vans, and SUVs were included. A total of 48 other vehicle exceedances was observed in about 20 h of observation. Typically, 1 to 3 other vehicles were being observed at any one time, and video analysis showed that this resulted in the observation of about 30 h of vehicle operation divided roughly into 80% freeway and 20% city streets. The observed exceedances were similarly divided. Dividing 48 by 30 gives about 1.6 exceedances/
vehicle/h as an observed baseline. In most cases it was not possible to determine what, if anything, caused the exceedance of the other vehicle. It could have been an in-vehicle device distraction; a personal reason; driver fatigue, inattention or impairment; or simply other random factors in the operation of the driver/vehicle/ roadway system. Exceedances were somewhat more frequent on curved sections of the freeways. Overall, this baseline rate and the rate observed in the destination entry trials were not particularly different for practical purposes. Vehicle speeds were recorded during the destination entry trials and familiarization runs. The drivers were instructed to drive at a comfortable, constant speed without exceeding the speed limit. On city streets, the average vehicle speed was 37 mph (59 km/h), slightly above the speed limit of 35 mph (56 km/h), with an overall speed range of about 74 mph (76 km/h). The average familiarization (no entry) vehicle speeds on city streets were similar. The average vehicle speed of 57 mph (91 km/h) on the freeway was below the speed limit of 65 mph (104 km/h), having a range of about 75 mph (78 km/h). The familiarization (no entry) vehicle speed on the freeway was higher than during the destination entry trials, by an average of about 3 mph (5 km/h), but the average familiarization speed was also under the speed limit. The extent to which freeway traffic may have influenced the average speeds was not analyzed. 5.3. Subjective rating results Each driver completed several rating forms after the 3 trials on the city streets, and another set after the 3 trials on the freeway. As a typical example, Fig. 17 shows the ratings for Ease of Street Name Entry (alphabetic keypad). Overall, this was rated between Easy and Moderately Easy for both city streets and the freeway.
Effortless
Easy Mode rate
Difficult Impossible City Streets Fig. 17. Ease of Street Name Entry.
Freeway
ARTICLE IN PRESS D.P. Chiang et al. / Applied Ergonomics 35 (2004) 215–223
223
in such areas as voice activation may further simplify the driver’s task or provide for the addition of other invehicle devices and information features, this study suggests that more conventional means of display and visual–manual data entry still represent an acceptable and effective means of interaction.
Effortless Easy Moderate Difficult
References
Impossible City Streets
Freeway
Fig. 18. Ease of Address Entry.
Fig. 18 shows the ratings for Ease of Address Entry (number keypad), which was rated Easy for both road types. Overall, the address was considered easier to input than the street name.
6. Discussion Overall, the results of this study show that drivers are able to make visual–manual destination entries with a contemporary navigation system while operating in actual traffic conditions; with acceptable levels of secondary task loading, path performance, and subjective lack of difficulty. The glance durations and eyes off road time behavior used by the drivers were within previously suggested rules of thumb for safe operation (e.g., Green, 1999; Tijerina, 1999, 2000; Farber, 2000; Greenberg; 2000; Zwahlen, 1988), and total entry times of 30–40 s were acceptable to the drivers when there was no need to enter the destination quickly. While future developments
AAM., 2002. Principles on Human Machine Interface (HMI) for InVehicle Information and Communication Systems (draft). Alliance of Automobile Manufacturers, Detroit. Farber, E., et al., 2000. Surrogate measures of visual demand while driving. Proceedings of the IEA 2000/HFES 2000 Congress, San Diego. Green, P., 1999. Visual and task demands of driver information systems, UMTRI 98-16. University of Michigan. Greenberg, J., 2000. Visual demand while driving, selected results from a Ford Motor Co. driving simulator study. Presented to the SAE ITS Safety and Human Factors Committee. Hashimoto, K., Atsumi, B., 2001. Study of occlusion techniques for making the static evaluation method of visual distraction. Presented at the WG8 Occlusion Workshop, Torino. ISO., 2002. Assessment of visual distraction, preliminary work item, ISO TC22/SC13/WG8, June 2002. SAE.m, 1999. Definition and experimental measures related to the specification of driver visual behavior using video based techniques. SAE J2396, Warrendale, PA. SAE., 2002. Navigation and route guidance function accessibility while driving (draft). Recommended Practice, SAE J2364, Warrendale, PA. Tijerina, L., 1999. Issues in the evaluation of driver distraction associated with in-vehicle information and telecommunication systems. SAE TOPTEC, San Diego, June 1999. Tijerina, L., et al., 2000. Driver distraction with wireless telecommunications and route guidance systems. DOT HS809-069. Zwahlen, H.T., et al., 1988. Safety Aspects of CRT Touch Panel Controls in Automobiles, Vision in Vehicles-II. Elsevier Science Publishers BV, Amsterdam.