Assessing the Safety of Infotainment Systems Used While Driving:

1
Assessing the Safety of Infotainment Systems Used
While Driving Practical Lessons from
InfoMan Paul Green Norimasa Kishi
University of Michigan UMTRI Ann Arbor, Michigan
48109-2150 USA Pagreen_at_umich.edu www.umich.edu/d
riving
Nissan Motor Company Nissan Research
Center Yokosuka, Kanagawa 237-8523,
Japan n-kishi_at_mail.nissan.co.jp
2
(No Transcript)
3
InfoMan (Information Manager)
Workload manager conceived and developed
by Nissan Proprietary workload algorithm
UMTRI provided a background literature
review to support its development UMTRI was a
collaborator in its evaluation
4

• What Did We Do? - Issues
• Experiment 1 (on the road, 2002 Q45)
• 1. Can an information manager reduce the risk of
  driving?
• 2. How can the Nissan workload metric be
  improved?
• Experiment 2 (DriveSafety driving simulator)
• What is the relationship between menu depth
  primary task demand?
• 2. Is PDT sensitive to workload?
• 3. How do the real and simulated results compare?

5
What Did We Do? - Driver Tasks
Hypothetical 4 /level real interfaces
Example 3 step Set the navigation heading to
north up. Example 6 step Set the map
background display to bright.
6
Simulated (Exp. 2)
Real (Exp. 1)
Arterial Downtown Rural M-14 (xway) US-23
(xway)
Arterial Downtown Rural M-14 (xway) US-23
(xway)
7
Lesson 1 - Simulated Roads
It is extremely difficult to match real roads
using the current version of the DriveSafety
software due to the tile architecture. No 2
lane/direction expressways Many missing x-way
interchanges No pipe bridges for signs
(landmarks) No easy way to add landmark
buildings
8
Specific Workload Roads - Experiment 2
9
Lesson 2 - Test Roads

• For large individual differences (e.g., young vs.
  old), variable length road sections are needed.
• --gt multiple out and back (return) points
• --gt hard to assemble consistent, multi-road
  sequence
• 2. Many driving studies occur in the
  spring/summer construction season. Get the
  schedules.
• 3. Because of traffic, data collection time-outs
  are needed.
• 4. Obtaining consistent data in urban areas is
  much more difficult than the open road, the
  typical case in the U.S.

10
Lesson 3 - Dep. Measures - New Scale
Rating important! SD of steering wheel
angle SD lane position (sim only) Headway (time
and distance) SD headway (sim only) Mean
speed Speed drops (sim only) Completion
time Response time Fraction of detected signals
Risk Driving Performance Menu
Task Peripheral Detection Task
11
Lesson 3 - New Rating Scale
Most use TLX to rate task workload/difficulty -
well researched - used by driving researchers
- has NASA name But Risk .ne. workload
Difficult to relate TLX to driving dimensions
(road geometry, traffic, etc.) Drivers need
contextual anchors for rating - range of
ratings is uncertain -gt range compression -
drivers forget what things were -gt inconsistency
12
Rating, As Risky As... 10. Driving with my eyes
closed. A crash will occur every time 9. Pass a
school bus red lights flashing the stop arm
out 8. Driving just under the legal alcohol
limit with observed weaving 7. 6. Driving 20
miles an hour faster than traffic on an
expressway 5. 4. Driving 10 miles an hour faster
than traffic on an expressway 3. 2. Driving on an
average road under average conditions 1. Driving
on an easy road with no traffic, pedestrians, or
animals while perfectly alert
(UMTRI Scale)
13
Lesson 3 - UMTRI Risk Scale Calibration
Extremely somewhat neither safe
somewhat extremelysafe safe nor unsafe
unsafe unsafe
1. Drive on an average road under average
conditions 2. Drive through a stop sign without
slowing down 6. Driving today Perform a
short task while driving on the hwy 7. Driving
today Perform a long task while driving on the
hwy Mark the risk ratings you gave while
driving (e.g., a 2). Below what value would
you perform a non-urgent task? Above what value
would you never perform in-vehicle tasks?
14
Lesson 3 - Key Points About Ratings
Most ratings are 2-6, so most ratings are
relative to speed (unidimensional). crash
prob. versus excess speed (for x-ways) -gt
fatalities (How many people will die
question.) Normalize to what is acceptable for
each driver so between driver comparisons are
consistent. (Different drivers have different
acceptance levels for excess speed.) Post-test
ratings (each task length (none, 3, 6 steps),
each road type) were well correlated with ratings
while driving. (-gt reliable)
15
Example Results from Risk Ratings
20 short tasks, 40 long tasks gt individual
safety thresholds
16
References Project reports to appear, see
www.umich.edu/driving Tsimhoni, Smith, and
Green, (2003). On-the-Road Assessment of Driving
Workload and Risk to Support the Development of
an Information Manager (Technical Report
UMTRI-2003-08) Tsimhoni, Smith, and Green,
(2003). The Effect of In-Vehicle Task Menu Depth
and Driving Workload on Task and Driving
Performance (Technical Report UMTRI-2003-09) Risk
Rating Boyle, Dienstfrey, and Sothoron (1998).
National Survey of Speeding and Other Unsafe
Driving Actions (NHTSA report)
17
Driver Distraction