paper presented at the Automotive Land Navigation Conference, June 18, 1997
`
`Potential Safety Impacts of Automotive Navigation Systems
`Paul Green
`University of Michigan Transportation Research Institute
`Human Factors Division
`Ann Arbor, Michigan 48109-2150 USA
`email: PAGreen@umich.edu
`
`Abstract
`
`This paper does not purport that navigation systems are safe or unsafe, but attempts to
`identify potential topics of concern relating to the driver interface. The first section of
`the paper describes some of the classes of measures by which the safety of navigation
`systems can be assessed. Of them, eye fixation frequencies and duration, TTC, and
`TLC have the most promise for pinpointing safety concerns. However, additional work
`is needed to develop the tools to collect and analyze such measures. Of the
`navigation tasks that drivers can perform, the primary concern is with destination entry,
`though information retrieval, route following (with maps), and destination retrieval also
`deserve attention. A better understanding of the crash-inducing potential of navigation
`systems may be obtained by consideration of likely crash scenarios, some of which
`are described. Systematic analyses of crash scenarios involving navigation systems
`have yet to appear in the literature.
`
`Introduction
`
`The purpose of this paper is to discuss (1) how the safety of a navigation system might
`be assessed experimentally, (2) the tasks drivers perform with navigation systems, and
`(3) the circumstances that could lead to crashes while using a navigation system.
`Readers should view this paper as an early version of "thought-piece" to encourage
`discussion, not as a scholarly review supported by empirical results or as a position
`paper for product liability actions.
`
`This information is presented from the perspective of a scientist who has done
`research on human factors issues associated with navigation systems. He also has a
`Rockwell Pathmaster navigation system installed in his own car and uses it
`periodically. The Pathmaster provides voice-assisted turn-by-turn guidance, has an
`arrow-based turn display, and can show route maps. The Pathmaster does not utilize
`traffic information to provide dynamic route guidance. The data base is reasonably
`complete and accurate except for the area to the west of where he lives. He does not
`drive in that area very often.
`
`How Can Safety Be Measured?
`
`Identifying the potential risks of a navigation system, if there are any, is extremely
`difficult. A first step in evaluating any traffic safety problem generally involves
`examining crash statistics. This can be a challenge for navigation systems as crashes
`in general are rare events. This, coupled with the low market penetration of navigation
`systems in the U.S. and Europe, results in too few cases (crashes) for a meaningful
`
`1
`
`Google Exhibit 1011
`
`Page 1 of 9
`
`

`
`statistical analysis. Even if there were enough cases, there is currently no record of
`whether a vehicle was fitted with a navigation system by the crash investigator, let
`alone whether the navigation system might have been in use. In the future, data on
`navigation system installation may be available (at least for OEM systems) by
`backtracking using the VIN. While the installation rate in Japan is quite high (soon to
`be in the millions of units), the Japanese National Police Agency, the primary source of
`accident data, tends to be quite restrictive in the information they release (in contrast to
`the western nations). The Japanese policy is changing. Nonetheless, crash-based
`insights into navigation system safety are unlikely in the near term.
`
`An alternative approach is to use secondary indicators as measures of crash risk
`(Burger, Smith, Queen, and Slack, 1977; Green, 1995a, b) . For example, in a recent
`UMTRI study participants used either an Ali-Scout or a Pathmaster navigation system
`to drive a four-segment route around Troy, Michigan (Katz, Fleming, Hunter, Green,
`and Damouth, 1996). While there were no critical incidents while driving using a
`PathMaster navigation system, there were four associated with the Ali-Scout (when
`subjects changed lanes without looking). (In each of these four crash opportunities,
`other drivers reacted by honking, braking, etc.) The problem in comparing the two
`interfaces is that there were over 50 Ali-Scout users but only 5 Pathmaster users (due
`to equipment problems). The study design did not permit collecting baseline data for
`using written instructions or paper maps. Thus, it is difficult to make safety inferences
`even from large and expensive studies. Future opportunities to conduct such research
`are rare.
`
`A third approach is to rely on more behavioral measures such as time-to-collision
`(TTC). Time-to-collision is defined as how long it would take for a collision to occur if
`all vehicles on the road retained their current velocity and acceleration tensors
`indefinitely. TTC is a measure of the safety cocoon around a vehicle, and
`consequently has considerable appeal for safety evaluations. However, devices for
`measuring TTC for safety evaluations are just being developed. Additionally, driver
`norms for TTC (needed to assess what is safe and unsafe) are lacking and there are
`no plans to develop norms.
`
`Even more indirect measures of crash potential are eye fixation, speed variance, and
`lane maintenance measures such as lane variance and time-to-line-crossing (TLC)
`(Godthelp, 1988; Godthelp, Milgram, and Blaauw, 1984). There is beginning to be a
`considerable body of data on glance durations. Glance data is usually collected by
`aiming a video camera at the driver's face, and playing back the tape at slow speed to
`determine the duration of each off-road glance and count their frequency. Data
`reduction time is typically 30-40 times subject testing times, so data reduction is often
`relegated to bored, low paid, undergraduate students. Although more automated
`methods are being developed, it will be several years before they are reliable and
`affordable enough for wide-spread use. As an aside, the technical challenges of
`automated systems include getting a magnetically-sensing head tracker to work in a
`magnetically-unfriendly environment (car body), counteracting the solar overload of
`the IR-based eye tracker, and resolving the optical interference of glasses, worn by
`virtually all older drivers. Further, norms on acceptable and unacceptable glance
`durations and frequencies from the road are limited (Hada, 1994).
`
`2
`
`Page 2 of 9
`
`

`
`Another approach would be to examine the influence of an installed navigation system
`on drivers' speed variance and lane variance. However, use of the navigation system
`is intermittent, so obtaining differences in normal driving on straight roads is difficult
`(Green, Hoekstra, and Williams, 1993; Green, Williams, Hoekstra, George, and Wen,
`1993). A solution to the intermittency problems is to partition each trip link into
`sections and only examine those sections where navigation system use is likely,
`namely near turns. (Walker, Alicandri, Sedney, and Roberts, 1990, 1991, 1992). The
`difficulty is that driving at the critical points is highly variable, so measuring differences
`is a challenge. This idea is discussed in greater detail in later portions of this paper.
`
`Not only are there problems in detecting differences, but it is uncertain what levels of
`performance are unacceptable as normative data on "plain old driving" are lacking.
`One interesting approach is to collect experimental data from drivers who are legally
`intoxicated and use that data to establish levels of unacceptable driving performance.
`The advantage of this approach is that is a large body of data relating BAC level to
`crash risk. This ideas assumes that the mechanism of impairment due to alcohol is the
`same as that due to other factors, which may not be the case.
`
`Thus, it will be some time before there is a clear picture of the safety consequences of
`using navigation systems while driving. The most useful insights may come from
`driver performance measures as measurement technologies evolve.
`
`What Tasks Do Drivers Perform with Navigation Systems?
`
`However, that does not mean that those interested in navigation system safety should
`do nothing while time passes. One option is to consider the conditions under which
`navigation system-induced crashes could occur and to target those conditions for
`examination. To avoid a biased perspective, the alternative--what would the driver do
`if a navigation system was not present--should also be considered. Would drivers be
`looking at a paper map while driving, a presentation scheme clearly less well suited to
`navigation than most contemporary navigation systems (Dingus, McGehee, Hulse,
`Jahns, Manakkal, Mollenhauer, and Fleischman, 1995)? Would drivers be looking a
`scraps of paper for directions? Would they slow to a crawl, trying to read
`nonilluminated street signs with 6-inch high letters at night? How likely are they to be
`rear-ended in this situation? Would they accumulate excess miles while searching for
`destinations, exposing themselves and others to increased crash risk by increasing
`the total vehicle miles traveled?
`
`Furthermore, stopping to use a navigation system (instead of continuing to drive) may
`be a risky option in some circumstances. For example, for most expressways, the
`safest option is to drive to the nearest exit and find a place to park. Uncertainties about
`the safety of the location is of major concern to drivers, and in some areas, just finding
`a place to park is not easy. If exits are widely spaced, a considerable distance may be
`traveled before the navigation information can be entered, eliminating wasted travel
`advantage of electronic navigation systems. The implications of stopping need to be
`explored more fully.
`
`There are four basic tasks drivers can perform with a navigation system: (1) enter a
`destination, (2) retrieve a destination, (3) obtain information about a potential
`
`3
`
`Page 3 of 9
`
`

`
`destination, (4) plan a route, and (5) follow guidance. The major concern is what a
`driver should be allowed to do while driving.
`
`Destination Entry
`
`Destination entry can be a challenging task, depending on the method used (Paelke
`and Green, 1993). Times on the order of a minute (in situations simulating a parked
`vehicle) are reported in the literature (Dingus, Hulse, Krage, Szczublewski, and Berry,
`1991; Steinfeld, Manes, Green, and Hunter, 1996). If the destination is entered using
`manual controls and visual display (presumably inside the vehicle), then the driver is
`distracted from attending to the road. Two possible solutions are offered. One, place
`the display on a HUD. While this solves the direction of gaze problem, it is unknown if
`this solves the locus of attention problem. This solution is worth investigating
`experimentally. Another alternative is to make entry by voice, as is the case for some
`systems in Japan. Published experimental evidence on voice entry is lacking, but
`desired.
`
`What a driver might enter varies quite widely with the interface style and features
`provided. Entry may require keying of every item, combinations of keying and
`scrolling, or scrolling entirely. Scrolling may involve scrolling though a list of items or
`scrolling a map to a particular location. Scrolling is a visually demanding task, a task
`not readily done while driving. On the other hand, scanning a keyboard to find one
`key among many may also have a high visual demand. Of particular concern is the
`"cognitive capture" issue, tasks that have a high start up cost (such as finding a
`reference point on an unstructured map), so that it is highly desired to complete them
`without interruption, even at the expense of other tasks. In driving, the result can be
`long periods of time without reference to the road.
`
`Entry tasks may require a driver to enter a street address, an intersection, an
`expressway exit, the destination or point of interest name, the destination phone
`number, or even its longitude and latitude. The author's experience is that phone
`number entry is much faster than other methods, but that method requires having
`current phone numbers in the data base. Destination entry times can be extremely
`large if the user does not know the name of the jurisdiction for the destination used by
`the navigation system. ("I know the street address and know where I want to go is just
`north of Ann Arbor, and I even know it is Washtenaw County, but I do not know the
`name of the city or township.") In these cases, several tries may be required. There is
`a reasonable probability the driver never obtains guidance to the exact street address
`desired.
`
`It is generally assumed that destination entry is done once per trip, usually before it
`starts. Often this is not the case. Some examples are:
`
`• The driver was in a hurry and knew the general direction in which to start. The
`destination was added immediately thereafter.
`• The driver decided to change destination enroute. ("Even though it is late, we should
`stop and get some milk on the way home." "Daddy, I need to go to the bathroom.")
`• The system does not use congestion information to calculate a route. Hearing a
`congestion report on the radio and having general knowledge of the area, the driver
`
`4
`
`Page 4 of 9
`
`

`
`planned the route around the congestion. When the congestion was passed, the
`destination was entered to obtain guidance for the final leg of the route to an
`unfamiliar destination.
`• The driver entered the wrong destination (North Main Street instead of South Main).
`• The driver does not know the exact destination at the beginning of the trip (e.g., the
`desired intersection) and therefore enters a location near the destination. Driving in
`the area near the interim destination provides information on the actual destination,
`which is then entered.
`
`Destination Retrieval
`
`Destination retrieval times in static situations are typically on the order of 10 seconds
`(Dingus, Hulse, Krage, Szczublewski, and Berry, 1991; Steinfeld, Manes, Green, and
`Hunter, 1996). Retrieval typically requires two keystrokes to get to a stored list, and
`then keying in the first few characters of a destination name or scrolling to get to it.
`Evidence as to whether lists should be alphabetical or by order of use (last first) is
`lacking.
`
`Route Following
`
`In route following, the driver is guided turn by turn, often by an arrow-like display
`(supplemented by voice) to the destination. The critical design feature is the timing of
`voice messages (George, Green, and Fleming, 1995; Green and George, 1995).
`Experience has shown that if the timing is off or the voice is too commanding (so
`drivers act without checking traffic), then problems will occur. Although visual displays
`help plan the route, execution of a turn seems to rely more on voice commands. A
`good voice system paired with a mediocre display may be acceptable, but the
`opposite is not true.
`
`The human factors data suggests that eye fixation times for navigation systems are of
`moderate duration except when maps are used (Dingus, McGehee, Hulse, Jahns,
`Manakkal, Mollenhauer, and Fleischman, 1995). In contrast to the U.S., maps are
`critical for driving in Japan because of differences in the nature of the Japanese road
`network. The consequence is more time spent looking away from the road. The major
`problems occur when drivers need to search a map for a street or place whose
`location on the map is completely unknown (as opposed to finding the street being
`driven or the next cross street). Times for scanning a map are close to those for
`destination retrieval.
`
`Information Retrieval
`
`At this time, navigation systems in the U.S. do not support information retrieval. In
`Japan, systems tend to be more information rich, and hence the use of the term
`"infomobile" instead of "automobile." Data on the potential tasks a driver might perform
`is lacking, but use of yellow page functions to obtain information on potential
`destinations in great detail (even the menu for a restaurant) is a possibility.
`
`5
`
`Page 5 of 9
`
`

`
`Administrative Functions
`
`This includes calibration and related activities. These activities are rarely performed
`and almost never at highway speeds, so safety implications of them are not a concern.
`
`When Might Navigation System-Related Crashes Occur?
`
`Thus, from the perspective of task duration, the primary tasks of concern are
`destination entry and information retrieval (tasks of potentially equal complexity). Map
`reading and destination retrieval are at a second level of concern. However, task
`frequency needs to be considered as well as duration. A destination is usually
`entered once per trip, but as was noted earlier, there are a few exceptions. Route
`guidance information is consulted numerous times on each trip.
`
`Following are a few navigation-related crash situations for both near and far from turn
`points. The first two involve destination entry. The second two involve route following.
`Additional thoughts as to the likelihood of these types of encounters would be useful in
`identifying potential navigation-related safety problems. The collision typology of
`Massie, Campbell, and Blower (1993) could be useful in this analysis.
`
`Situation 1 : The driver is on an expressway and looking at the navigation system while
`entering a destination and
`
`• something occurs without warning requiring an immediate response (lead vehicle
`brakes or another vehicle cuts in).
`• the driver drifts out of the lane into another vehicle.
`• the road curves but the driver continues to steer straight.
`• the driver does not see a lane drop and drives off the road.
`• the driver does not see an obstruction (parked vehicle) and collides.
`• there is a general change in the speed of traffic (due to a sag in the road) and the
`driver collides with a lead vehicle.
`• an animal darts out on to the road and the driver never sees it.
`
`Situation 2 : The driver is in an urban area and looking at the navigation system while
`entering a destination and
`
`• the driver misses a stop sign or traffic light and collides with a vehicle on a crossing
`path. This might include never seeing the signal at all or not noticing a state change.
`• traffic ahead slows (for a variety of reasons) and the driver rear-ends a lead vehicle.
`• a parked car from the roadside or a driveway partially obstructs the driver's path but
`the driver never swerves to avoid it.
`
`Situation 3 : The driver is on an limited-access road and receives guidance too late,
`either because the guidance was poorly timed or the driver missed the prepare-to-exit
`message. The driver hastily attempts to change lanes to get to the exit.
`
`6
`
`Page 6 of 9
`
`

`
`Situation 4 : The driver is in an urban area and
`
`• receives the turn message too late and brakes abruptly, inviting a potential rear end
`collision.
`• receives a prepare-to-turn or turn message. Thinking it to be a command (and
`believing the computer knows all) the driver turns or changes lanes without checking
`for traffic.
`• receives a prepare-to-turn or turn message. Thinking it to be a command, the driver
`turns down a one-way street in the wrong direction. This is a data base-induced
`error.
`• receives a prepare-to-turn or turn message. Thinking it to be a command, the driver
`ignores a traffic sign or signal.
`• receives a prepare-to-turn or turn message, but is unsure of the exact turn location.
`The driver proceeds at an excessively slow speed and is rear ended.
`• receives a prepare-to-turn or turn message, but is unsure of the exact turn location.
`The driver is so intent on finding the turn point that they miss a traffic sign or signal
`and strike a crossing vehicle.
`
`Closing Remarks
`
`Thus, there are technological means to determine if there are safety problems
`associated with navigation systems. However, it will be sometime before they are
`developed to the level that safety questions are easy to examine. In the interim, in
`addition to pursuing the development of those technologies aggressively, some
`thought needs to be given to the circumstances under which navigation-induced
`crashes could occur. Using this approach, areas of concern could be identified and
`navigation interfaces could be designed so that even hints of a crash-induced
`potential could be removed. The success of the navigation market in the U.S. and
`Europe will depend on reducing the system cost to a more reasonable level, and as
`indicated by this paper, enhancing the safety (and, consequently the usability) of driver
`interfaces.
`
`References
`
`Burger, W.J., Smith, R.L., Queen, J.E. and Slack, G.B. (1977). Accident and Near
`Accident Causation: The Contribution of Automobile Design Characteristics
`(technical report DOT-HS-802-714). Washington, D.C.: U.S. Department of
`Transportation, National Highway Traffic Safety Administration.
`
`Dingus, T., McGehee, D., Hulse, M., Jahns, S., Manakkal, N., Mollenhauer, M., and
`Fleischman, R. (1995). TravTek Evaluation Task C3 - Camera Car Study (Technical
`Report FHWA-RD-94-076), McLean, VA: U. S. Department of Transportation,
`Federal Highway Administration.
`
`Dingus, T.A., Hulse, M.C., Krage, M.K., Szczublewski, F.E., and Berry, P. (1991). A
`Usability Evaluation of Navigation and Information System "Pre-Drive" Functions
`(SAE paper 912794), VNIS'91 Proceedings , Warrendale, PA: Society of Automotive
`Engineers, 527-536.
`
`7
`
`Page 7 of 9
`
`

`
`George, K., Green, P., and Fleming, J. (1995). Timing of Auditory Route Guidance
`Instructions , (Technical Report UMTRI-95-6), Ann Arbor, MI: The University of
`Michigan Transportation Research Institute.
`
`Godthelp, H. (1988). The Limits of Path Error-Neglecting in Straight Line Driving,
`Ergonomics , 31 (4), 609-619.
`
`Godthelp, H., Milgram, P., and Blaauw, G.J. (1984). The Development of a Time-
`Related Measure to Describe Driving Strategy, Human Factors, 26 (3), 257-268.
`
`Green, P. (1995a). Measures and Methods Used to Assess the Safety and Usability of
`Driver Information Systems , (Technical Report FHWA-RD-94-088), McLean, VA:
`U.S. Department of Transportation, Federal Highway Administration.
`
`Green, P. (1995b). Suggested Procedures and Acceptance Limits for Assessing the
`Safety and Ease of Use of Driver Information Systems (Technical Report FHWA-RD-
`94-089), McLean, VA: U.S. Department of Transportation, Federal Highway
`Administration.
`
`Green, P. and George, K. (1995). When Should Auditory Guidance Systems Tell
`Drivers to Turn?, Proceedings of the Human Factors and Ergonomics Society 39th
`Annual Meeting , Santa Monica, CA: Human Factors and Ergonomics Society, 1072-
`1076.
`
`Green, P., Hoekstra, E., and Williams, M. (1993) On-The-Road Tests of Driver
`Interfaces: Examination of a Navigation System and a Car Phone (Technical Report
`UMTRI-93-35), Ann Arbor, MI: The University of Michigan Transportation Research
`Institute.
`
`Green, P., Levison, W., Paelke, G., and Serafin, C. (1995). Preliminary Human Factors
`Guidelines for Driver Information Systems (Technical Report FHWA-RD-94-087),
`McLean, VA: U.S. Department of Transportation, Federal Highway Administration.
`
`Green, P., Williams, M., Hoekstra, E., George, K. and Wen, C. (1993). Initial On-the-
`Road Tests of Driver Information System Interfaces: Examination of Navigation,
`Traffic Information, IVSAWS, and Vehicle Monitoring (Technical report UMTRI-93-
`32), Ann Arbor, MI: The University of Michigan Transportation Research Institute.
`
`Hada, H. (1994). Drivers' Visual Attention to In-vehicle Displays: Effects of Display
`Location and Road Types (Technical Report UMTRI-94-9), Ann Arbor, MI: University
`of Michigan Transportation Research Institute.
`
`Katz, S., Fleming, J., Hunter, D.R., Green, P., and Damouth, D. (1996). On-the-Road
`Human Factors Evaluation of the Ali-Scout Navigation System (Technical Report
`UMTRI-96-32), Ann Arbor, MI: University of Michigan Transportation Research
`Institute.
`
`8
`
`Page 8 of 9
`
`

`
`Massie, D.L., Campbell, K.L., and Blower, D.F. (1993). Development of a Collision
`Typology for Evaluation of Collision Avoidance Strategies, Accident Analysis and
`Prevention , 25 (3), 241-257.
`
`Noy, I. (1990). Attention and Performance While Driving with Auxiliary In-Vehicle
`Displays (Transport Canada Publication TP 10727 (E)), Ottawa, Ontario, Canada:
`Transport Canada, Traffic Safety Standards and Research, Ergonomics Division.
`
`Paelke, G. and Green, P. (1993). Entry of Destinations into Route Guidance Systems:
`A Human Factors Evaluation (Technical Report UMTRI-93-45), Ann Arbor, MI: The
`University of Michigan Transportation Research Institute.
`
`Steinfeld, A., Manes, D., Green, P., and Hunter, D. (1996). Destination Entry and
`Retrieval with the Ali-Scout Navigation System (Technical Report UMTRI-96-30),
`Ann Arbor, MI: The University of Michigan Transportation Research Institute.
`
`Van Winsum, W. and Heino, A. (1996). Choice of Time-Headway and the Role of
`Time-to-Collision Information in Braking, Ergonomics , 39 (4), 579-592.
`
`Walker, J. Alicandri, E., Sedney, C., and Roberts, K. (1991). In-Vehicle Navigation
`Devices: Effects on the Safety of Driver Performance, Vehicle Navigation and
`Information Systems (VNIS'91) , New York: Institute of Electrical and Electronics
`Engineers, pp. 499-525.
`
`Walker, J., Alicandri, E., Sedney, C., and Roberts, K. (1990, May). In-Vehicle
`Navigation Devices: Effects on the Safety of Driver Performance (FHWA/RD-90-
`053), McLean, VA: Federal Highway Administration, U.S. Department of
`Transportation.
`
`Walker, J., Alicandri, E., Sedney, C., and Roberts, K. (1992). In-Vehicle Navigation
`Devices: Effects on the Safety of Driver Performance, Public Roads , 56 (1), pp. 9-22.
`
`9
`
`Page 9 of 9