embedded systems. ⢠Real-time Steering Entropy (RSE) has been proposed as a feasible approach .... to known long distance drivers: Truck, Coach,. Rideshare.
Preliminary Evaluation of Real-time Steering Entropy for Monitoring Driver State Takayuki Kondoh (Nissan Motor, VTTI) Shane McLaughlin (VTTI) Tomohiro Yamamura (Nissan Motor) Nobuyuki Kuge (Nissan Motor) Miguel Perez (VTTI) Takashi Sunda (Nissan Motor)
Contents • Background • Purpose • Real-time Steering Entropy(RSE) • Naturalistic Driving Study and Previous Study • Results • Conclusion
www.nissan-global.com
Contents • Background • Purpose • Real-time Steering Entropy(RSE) • Naturalistic Driving Study and Previous Study • Results • Conclusion
www.nissan-global.com
Background • Distraction and fatigue are of course contributing factors in vehicle crashes • Countermeasure: driver state monitor • camera based • physiological based • driving behavior based
• These can challenge computational resources of embedded systems • Real-time Steering Entropy (RSE) has been proposed as a feasible approach www.nissan-global.com
Contents • Background • Purpose • Real-time Steering Entropy(RSE) • Naturalistic Driving Study and Previous Study • Results • Conclusion
www.nissan-global.com
Purpose • Two previous studies have clarified as follows: – The RSE method was demonstrated to detect various forms of driver distractions – The RSE method was found to be sensitive to calling behavior (incoming call and outgoing calls)
RSE performance was inves2gated for use in iden2fying periods of reduced alertness 1) in long-dura2on driving situa2ons, and 2) at different 2mes during the day www.nissan-global.com
Contents • Background • Purpose • Real-time Steering Entropy(RSE) • Naturalistic Driving Study and Previous Study • Results • Conclusion
www.nissan-global.com
Real-time Steering Entropy(RSE) Rationale for Measurement • Steering behavior is generally smooth for a driver who is concentrating on driving PE (diff. btwn “actual steering angle” and “predict”) is small PE appears at the center on zero in P.D. of PE
• Quick steering inputs indicate steering corrections and high workload (ex. after looking aside) ⇒ PE gets great. ⇒ Distribution of PE widens: Hp (measure of disorder) Usual increases After Looking aside
L greater www.nissan-global.com
0
R greater
Real-time Steering Entropy(RSE) Measurement 1. Uses a comparison instantaneous measures to long-term measures (driver characteristics) 2. RSE computation needs time to learn and identify individual steering characteristics 3. Relative entropy is applied to quantify both measures.
www.nissan-global.com
Real-time Steering Entropy(RSE) RHp Behavior in Driving • Examples of changes in one trip
www.nissan-global.com
Contents • Background • Purpose • Real-time Steering Entropy(RSE) • Naturalistic Driving Study and Previous Study • Results • Conclusion
www.nissan-global.com
NDS and Previous Study NDS Summary • Collected in collaboration with VTTI • Approval(#10-624) from IRB of VT to analysis
OVERALL Driving Trip: 2,303 Driving Distance: 39,000km Driving Time: 730hours
Face
Forward
Hand and Instrument Panel
Top: Rear / Bottom: Foot
Participants: 18
Extracted Trips Over 60 kph At least For 15 min
www.nissan-global.com
Driving Trip: 263 Driving Distance: 10,500km Driving Time: 190hours
RSE
NDS and Previous Study Observed Driver Condition • 30% (64 events): Reduced alertness • 39% (83 events): Cell phone, Talking to passenger • 22% (46 events): Looking aside, Reaching for something to drink/eat One Hand on wheel No hands on wheel Irritation, 2% 1% 4% Touching Body, 3% Drinking/Eating, 3%
Reduced alertness
Reaching for something to drink/eat, 4% Operating navigation controls, 4%
30% 64 events
Looking aside, 8%
Singing, 1%
Talking, 13%
Tired, 1%
Calling , 26%
www.nissan-global.com
NDS and Previous Study Typical RHp behavior in Calling Events • RHp appears to follow onset of distraction – Outgoing call: Increases before the call started – Incoming call: Increases after the call started Outgoing call
The call started
Incoming call
The call started
prepare to take phone, dialing
pressure receiving ASAP after ringing
www.nissan-global.com
Contents • Background • Purpose • Real-time Steering Entropy(RSE) • Naturalistic Driving Study and Previous Study • Results • Conclusion
www.nissan-global.com
Results Summary of Reduced alertness • Reduced alertness was found for 30% of events • This is the largest category of events One Hand on wheel No hands on wheel Irritation, 2% 1% 4% Touching Body, 3%
• Additional information was collected for each of these events • elapsed time • time of day
Drinking/Eating, 3% Reaching for something to drink/eat, 4% Operating navigation controls, 4%
Reduced alertness
30% Looking aside, 8%
64 events
Singing, 1%
Talking, 13%
www.nissan-global.com
Tired, 1%
Calling , 26%
Results Elapsed time • Longer driving leads to higher possibilities of drowsiness 100% High-RHp events
90%
Reduced alertness events involve on high-Rhp events
80%
Share (%)
70% 60% 50% 40% 30% 20% 10%
38 27 21 16 16 12 10 9 6 6 1 1 1
0% 0
0.5
1
1.5
2 2.5 3 Elapsed Time (h)
www.nissan-global.com
3.5
4
4.5
5
Results Time of Day • The peaks of reduced alertness appear cyclical (~1.5~2 hours), even though the rate of overall is constant 60% High-RHp events Reduced alertness events involve on high-RHp events
50%
Share (%)
40%
30%
20%
10%
0%
10 16 15 13 14 14 19 15 11 14 23 20 17 15 6 4 4 13 29 8 8 2 2 6 7 2 3 6 7 2 0 7 6 96 54 44 19 1 1
2
3
4
5
6
7
8
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Time of Day (h)
www.nissan-global.com
Results Discussion • Longer periods of driving correspond with more frequent periods of reduced alertness. • The peaks in percentage of high-RHp events where reduced alertness was observed were found to be approximately three hours apart. – It may show “Ultradian Rhythm”
www.nissan-global.com
Contents • Background • Purpose • Real-time Steering Entropy(RSE) • Naturalistic Driving Study and Previous Study • Results • Conclusion
www.nissan-global.com
Conclusion • Applying RHp to NDS was successful in finding periods of reduced alertness. • These periods appeared to be 1) in long-duration driving situations 2) at different times during the day RSE approach has the potential to reliably detect higher-risk driving conditions
www.nissan-global.com
Thank you for your attention
Possibilities of RSE - A good index of various driver states and even phases of tasks (e.g., sensitivities related to cell use) - Good for processing big databases: SHRP2 etc - Application to known long distance drivers: Truck, Coach, Rideshare www.nissan-global.com
