automotive ui 2011
DESCRIPTION
Reducing Driver DistractionIn-Car Communication SystemDriving Situations MonitoringTRANSCRIPT
The Impact of and Adaptive User Interface
on Reducing Driver Distraction
Authors: Patrick Tchankue, Janet Wesson and Dieter Vogts
3rd International Conference on Automotive User Interface,
November 29-December 2, 2011, Salzburg, Austria
Overview
• Background
• In-Car Communication Systems
• Driver Distraction
• Adaptive Interfaces
• Architecture of MIMI
• User Study
• Results
• Conclusion & Future work
Background
• In-Car Infotainment Systems are becoming common
– Information: communication, navigation and safety;
– Entertainment: radio, CD and games;
– Hands-free and eyes-free: voice-activated;
• Existing UI were not initially designed for such
applications
In-Car Communication Systems (ICCS)
• Most common component of in-car systems:
– Manage calls, text messages and contacts in the car via
Bluetooth (hands-free);
– Use speech (eyes-free) and steering wheels buttons (hands-
free) as input channel.
• Examples of ICCS:
Name Manufacturer Year
iDrive BMW 2001
Blue&Me Fiat 2004
SYNC Ford 2007
IQon SAAB 2011
Driver Distraction
• Driver distraction occurs when the driver’s attention is
diverted from driving to the extent that the driver is no
longer able to drive adequately or safely (Young &
Regan, 2005).
• Type of driver distraction:
– Visual: taking your eyes off the road;
– Auditory: internal and external noises;
– Manual: taking your hands off the steering wheel; and
– Cognitive: taking your mind off what you’re doing.
• Texting can cause more serious driver distraction.
Adaptive Interfaces
• Interfaces able to adapt to specific user, task or
situations;
• Inferring the distraction level;
– Fuzzy logic;
– Support Vector Machine;
– Neural networks;
• Adaptation effects
– Delaying calls and text messages;
– Resuming the notification process;
– Warn drivers before potential dangerous outgoing events
Input Module (A)
Architecture of MIMI
NL Understanding
Multimodal Fusion
ASR
Dia
log
ue
ma
na
ge
r (C
)
Mobile phone
Adaptive module
Inputs
Dialogue
history
Task
progress
Workload
manager
Knowledge base
User
model
Task
model
Context
model
Adaptive engine
Mobile phone
interface (B)
Output Module (D)
CAN bus
Phonebook DB
NL
generation TTS
Dialogue engine
Architecture of MIMI (cont.)
• Workload manager
speed
Δ speed
angle
Δ angle
Distraction
level
1 = very low
3 = mid
5 = very high
2 = low
4 = high
User Study
• Aim
– Usability (task success, errors, effectiveness of tasks, time of
task)
– Safety (cognitive load, mean lateral deviation, perceived safety,
adaptation)
• Methodology
• Participants
– 30 students
• Tasks
– Calling
– Sending text messages
Results (cont.)
• Usability
Comparison of the usability of the non-adaptive and adaptive version of MIMI (n=30)
6.10
6.23 5.73
5.43
4.47
6.17
6.33 5.90
5.70
5.07
1
2
3
4
5
6
7
Call effectiveness SMS effectiveness Barge-in Recognition Number dictation
Non adaptive Adaptive
Results (cont.)
• Performance
Comparing the mean time-on-task (in seconds) for MIMI 1 and MIMI 2 (n=30).
MIMI 1 non adaptive MIMI 2 adaptive T-test
Mean StdDev Mean StdDev p-value
T01 11 19.02 14.5 19.98 0.16
T02 10 5.22 21.5 7.34 0.00
T03 36 34.8 40.5 34.47 0.78
T04 11 6.28 20.5 7.56 0.00
T05 11 12.75 10.5 8.79 0.88
T06 10 4.29 18 7.45 0.00
T07 23 12.09 22 12.57 0.78
T08 11 9.95 20.5 18.64 0.05
T09 7 10.91 7 11.01 0.97
T10 10 5.34 16.5 38.75 0.08
Results (cont.)
• Performance
MIMI 1 non adaptive MIMI 2 adaptive T-test
Mean StdDev Mean StdDev p-value
T01 1.16 0.7 0.98 0.69 0.21
T02 0.94 0.49 0.87 0.43 0.18
T03 1.86 0.58 1.77 0.51 0.33
T04 0.99 0.45 0.95 0.43 0.68
T05 1.25 0.49 1.24 0.45 0.78
T06 1.05 0.56 0.84 0.53 0.52
T07 1.62 0.56 1.43 0.64 0.14
T08 1.13 0.5 1.02 0.4 0.28
T09 1.05 0.71 1.1 0.51 0.65
T10 0.96 0.63 0.79 0.48 0.47
Comparing the mean lateral deviation (in meters) for MIMI 1 and MIMI 2 (n=30)
Results
• Safety
5.13
5.43 5.37 5.63
5.43
5.87 5.97 6.17
1
2
3
4
5
6
7
Safe to make calls Safe to send SMS Safe to answer calls Safe to read SMS
Non adaptive Adaptive
Comparison of the safety ratings of non-adaptive versus the adaptive version of MIMI (n=30).
Results
• Adaptation
Comparing the adaptation of MIMI 1 and MIMI 2 (n=30).
Postponing Warning sound
MIMI 1 non
adaptive
MIMI 2
adaptive
MIMI 1 non
adaptive
MIMI 2
adaptive
Mean 4.76 5.80 4.80 4.80
Median 5.00 6.00 4.00 5.00
Mode 4.00 6.00 4.00 4.00
StdDev 1.87 1.45 1.56 1.65
p-value 0.01 1.00
Conclusion & Future work
• ICCS can be affected by usability and safety issues;
• An adaptive interface for an ICCS was designed;
• A user study compared MIMI 1 and MIMI 2 in terms of
usability and safety;
• The Adaptive interface had a positive impact on the
usability and safety of MIMI;
• Future work
– Other adaptation effects to be investigated;
– Alternative warning strategies.
Questions ?
Contact: Emails: [email protected]
Website: www.nmmu.ac.za/cs
Tel: +27 41 504 2323
Thank you for your attention!