session 50josef nilsson
TRANSCRIPT
Föraragerande och säkerhetseffekter vid
felfunktion i ACC-systemet – resultat från ensimulatorstudie
13 Jan 2011
Josef Nilsson*¥, Niklas Strand*Ŧ
* SAFER Vehicle and Traffic Safety Centre ¥ SP Technical Research Institute of Sweden
Ŧ VTI Swedish National Road and Transport Research Institute
Acknowledgements
SHADES - System safety through combination of HMI And Dependable Systems (to improve driver assistance systems)
Purpose of the study
Experimental setup
– Failure situations
Experiences with the chosen setup
Preliminary results
Outline
Work in progress!
What are the safety effects of technical failures of an adaptive cruise control system (ACC)?
What experimental setup should be used for this type study?
Purpose
Adaptive Cruise Control (ACC)
Four failure modes
– Two acceleration failures
– Two brake failures
All with the same initial setting:
– ACC activated
– 105 kph
– Following leader with a 2 second time-gap
– No vehicle in left lane (free to overtake)
Experimental setup – Failure situations
Ego
Lead
a. Car in front drives at 105 kph, ACC in the ego car accelerates towards vehicle ahead
b. Car in front brakes, ACC in the ego car does not brake
c. Car in front brakes, ACC in the ego car brakes but with less force than what is required to avoid a collision
d. Car in front accelerates from 105 kph to 150 kph, ACC in the ego car keeps the set distance but fails to keep the set speed limit (110 kph)
Baseline (a, b, c, and d): Same situations but no failure
Experimental setup – Failure situations
Chalmers Driving Simulator
Moving-base (hexapod)
VTI software– Graphics, Sound
– Executing scenario
– Logging
Fault injection support– Fault injection interface to simulink
models of ACC
– Fault triggered from scenario
Automatic safety stop if vehicle becomes unstable
Experimental setup
Highway (110 kph limit)
Summer conditions
Low traffic density
At least 5 minutes driving before each situation
48 subjects (23 men, 15 women)
Between 25 and 60 years of age
Drivers license for more than 5 years
Annual mileage more than 5000 km
No experienced ACC users
Experimental setup - Sample
1. Written instructions, Consent form
2. Survey (background + locus of control)
3. Further instructions and Training
4. Drive/baseline (no failure) Drive/situation (failure)
5. Drive/baseline (no failure) Drive/situation (failure)
6. Interview
7. Survey (simulator realism/driving quality)
Step 4 and 5 are counterbalanced (2 scenarios + 2 baselines)
Experimental setup - Procedure
Brake situations (leader brakes but ego vehicle fails to brake correctly):
Drivers are prepared for braking leaders and changes lane at first sign of leader decelerating
– Left lane is free of traffic
– Each failure situation preceded by baseline
– Some subjects drive two brake failure scenarios
Between subject design perhaps more appropriate
Difficulties
All scenarios:
Failure not triggered (ego vehicle never following leader) (7 subjects)
– Drivers keep overtaking since left lane is free of traffic most of the time
Adding overtaking vehicles not always helpful
Difficulties (cont.)
Failures go unnoticed despite high criticality
– Drivers immediately overtake when braking fails
– Keeps driving with latent fault in the ACC
Drivers do nothing when vehicle accelerates
Observations
10 collisions
– No collisions from acceleration failures
– 3 collision for complete loss of braking
– 7 collisions for partial loss of braking
5 safety stops (unstable vehicle)
– No safety stops for brake failures
– 2 safety stops when ACC is speeding but still keeping 2 s distance to leader
– 3 safety stops when ACC is speeding
Observations (cont.)
Developed a fault injection platform for driving simulator experiments
Difficulties in experiment design
– Problematic to keep driver following leader while simultaneously allowing overtaking
– Brake failures to alike – Within-subject design problematic
Preliminary results show:
– Partial brake failures led to most collision
– Several acceleration failures led to instability
No action from some subjects when ACC was speeding
Summary
Thank you!
E-mail: [email protected], [email protected]
Dependent variables
Type of strategy
– Steering
– Braking
– ACC off button
– Combination
Outcome of the situations
Response time
Minimum time-to-collision
Time until time-gap recovered