Post on 13-Jul-2015
Experiment with sleepy drivers on real road
» Autoliv» Chalmers» VTI» VCC
Aim» Test of feedback and warning strategies during
the development of sleepiness based on algorithm and/or subjective ratings» How effective?» Driver acceptance?» Lasting effects?
» Evaluation of existing algorithm
Driving procedurer & participants» 12 persons (2 persons each 24 hour-
period) » Random sample from national register of
vehicle owners» Volvo S80 – double command, test leader» Road 34 from Linköping to south of Kisa» Each person drove ~140 km five times
(dose – response)» Arrive at VTI 8.00. (information, electrodes,
questionnaire, …)» Time between test drives was regulated» Finish 03.00 and 05.00 respectively» Test 15 – 30 October 2007
Driving time subject A
Driving time subject B
09.00 11.00
13.00 15.00
17.00 19.00
21.00 23.00
01.00 03.00
Drowsiness detection» Algorithm
» Lane position» Steering wheel angle» Speed» Yaw angle» Accelerometer
» References» EEG, EOG, EMG, EKG» DSS (Driver State Sensor)
» PERCLOS, Blink duration etc» Video recording
» Trigger» A recommendation was given to
the test leader from the algorithm and confirmed / send to the participant manually
Feedback and warning strategy
tIncidentDiagnos 1 Diagnos 2
(Kritiskt läge)
Index från algoritm
Δt = tillräcklig tid för åtgärd
Pigg
Sover
LDW
FCW
1
2
3
x
DROWSI – warning strategy
(1:a iden.) (2:a = D1)(D2)
1. Feedback
2. Warning 1
Arousal (”pling”)
Visual (text message)
3.Warning 2
Arousal (”pling-pling”)
Vibrations
-Steering wheel
-Seat belt
Visual (text message)
Data» Vehicle data (CAN bus + external sensors)» Physiological data (EEG/EOG/EMG + Camera (SE))» Subjective sleepiness (5 minute)
Preliminary results
» KSS» Drivers opinion about the warning strategy
and modalities» Effects on driver sleepiness and driving
behaviour
Subject 1-12
KSS
Feedback 9 participants reported experience from feedback
Did the feedback...
8
56
1
32
01 1
0
2
4
6
8
10
12
14
16
18
felt useful? (n=9) motivated you to dosomething aboutyour sleepiness?
(n=9)
influenced yourdriving performance?
(n=9)
num
ber o
f driv
ers
YesNoDon't know
6 felt more awake3 were surprised
FeedbackThe feedback, was it...
0 0
2
4
1 1
6
3
01
01
87
0 00
2
4
6
8
10
12
14
16
18
annoying? frightening? obvious why thenumber of columns
changed?
obvious what to dowhen the number of
columnsincreased?
num
ber o
f driv
ers
YesYes, fairlyNo, doubtfullyNo
Warning 1 9 participants reported experience from warning 1
Did the warning...
7
9
100 0
0
2
4
6
8
10
12
motivated you to do somethingabout your sleepiness?
influenced your drivingperformance?
num
ber o
f driv
ers
YesNo Don't know
8 felt more awake3 were surprised
Warning 24 participants reported experience from warning 2
Did the warning influence your driving?
3
11
3
0 001234
Improved your alertness(n=4)
surpriced you? (n=4)num
ber o
f driv
ers
Yes
No
Don't know
ConclusionVisual displayThe idea with feedback is experienced as positive, but it was not clear how to understand it.
Suggestion: Invert the bars
Vibrations
Effective and useful – but need to have a high amplitude
The seat belt vibrations were experienced as most annoying and most frightening.
Truck experiment - Volvo
+ sound
+ sound
+ vibrations in seat belt
Photo: Katja Kircher
Effects on driver sleepiness and driving behaviour
Participants» Feedback - 5 participants out of 9 possible to use for
sdlp, 7 for KDSmax» Warning 1 – 6 participants out of 9 possible to use» Warning 2 – not enough data
Effects in terms of–Sleepiness indicator – Karolinska Drowsiness score (KDS) & blink duration–Driving behaviour – variability in lp (sdlp)
Feedbacksdlp
0.000.050.100.150.200.250.300.350.40
Reference (min 1-5) Before (min 5-1) After (min 1-5)
Met
ers
No sign. differences within reference period, before period or after periodNo sign. Differences between before period or after period
FeedbackKDS max
0
5
10
15
20
25
Reference (min 1-5) Before (min 5-1) After (min 1-5)
Perc
ent
No sign. differences within reference period, before period or after periodNo sign. Differences between before period or after period
Warning 1sdlp
0.00
0.05
0.10
0.15
0.20
0.25
0.30
Reference (min 1-5) Before (min 5-1) After (min 1-5)
met
er
No sign. Difference between Ref. minute 5, Before minute 1 & After minute 1
No sign. differences within reference period, before period or after period
Warning 1KDSmax
048
121620242832
Reference (min 1-5) Before (min 5-1) After (min 1-5)
Perc
ent
No sign. Difference between Ref. minute 5, Before minute 1 & After minute 1
No sign. differences within reference period, before period or after period
ConclusionDriving behaviour (sdlp)
» No significant effects of feedback» No significant effects of warning 1
Sleepiness indicators (KDSmax)» No significant effects of feedback» No significant effects of warning 1
MethodMajor problems with data loss from lane tracker, artefacts in physiological
measures (EEG/EOG)High degree of ecological validityHigh degree of control of subjects Low degree of environmental control (road section, weather, traffic…)Individual differences more participants needed