potential friction - IPG Automotive GmbH

Date: 2012, Sept. 18
Potential friction knowledge benefits on
ADAS PCS-ACC systems:
logic, performances and simulated
critical scenarios.
AUTHORS:
Soluzioni Ingegneria
Elisabetta Leo
Marco E.Pezzola
Politecnico di Milano
Federico Mancosu
Federico Cheli
ADAS braking system world
SAFETY
TARGET: to investigate
the probability of an
accident
TARGET: impact
mitigation
MONITORING driver:
• steering input
• braking/throttle
• internal cameras
RESEARCH TARGET:
to verify benefits of tyre-road condition
knowledge to avoid accident
TARGET: to avoid
accident
ADDING:
• Radar/lidar/cameras
• V2V communication
• V2I communication
• Tyre-road condition
• Vehicle dynamic
• ….
POTENTIAL FRICTION IDENTIFIER v3.2
18/09/2012
N°2
Potential friction identification
Higher longitudinal slip
Constant speed
Accelerating/braking
Higher lateral slip
Longitudinal maneuvers
Lateral maneuvers
TARGET: to identify the potential friction before reaching it. As
before as possible!
18/09/2012
N°3
Index
1. ADAS braking system logic description (including potential
friction - 0 - infos).
2. Evaluation of 0 knowledge benefits (VIA CarMaker
simulations): critical scenarios’ definition.
3. “ 0 identification algorithm” (longitudinal transient):
experimental tests and results.
4. Implementation of the “ 0 identification algorithm” in
CarMaker & Real-Time analysis.
5. Conclusions.
18/09/2012
N°4
Index
1. ADAS braking system logic description (including potential
friction - 0 - infos).
2. Evaluation of 0 knowledge benefits (VIA CarMaker
simulations): critical scenarios’ definition.
3. “ 0 identification algorithm” (longitudinal transient):
experimental tests and results.
4. Implementation of the “ 0 identification algorithm” in
CarMaker & Real-Time analysis.
5. Conclusions.
18/09/2012
N°5
ADAS braking system logic description
6
Vehicle speed (i.e. via CAN bus)
µ: potential friction
(direct or indirect mode)
Relative speed and
distance between
vehicles: radar
Index 1
SAFETY
INDEXES
COMPUTATION
Index 2
CONTROL MODE
definition
1) SAFE
2) WARNING
3) DANGER
Desired
acceleration
REAL or SIMULATED WORLD
% throttle
% Brake
CONTROL
LOGIC
18/09/2012
N°6
ADAS braking system logic description
INDEX 1: “1/TTC”
INDEX 2: “Ad_D”
Nomenclature
TTC = Time to collision
Ad_D = Adimensional Distance.
Description:
Time to reach the in-front-vehicle
assuming no driver actions
Margin between required distance to stop
the vehicle before collision and actual
relative distance
Function of:
Relative distance
Relative speed
Relative speed
Relative distance
Relative distance
Relative speed
Potential friction
Human’s time response to activate brake
Vehicle dynamics
Safety if
TTC: HIGH  1/TTC: LOW
Ad_D: HIGH
18/09/2012
N°7
f(Time To Collision)
ADAS braking system logic description
DANGER
maximum
deceleration
WARNING
high deceleration
allowed
WARNING
high deceleration
allowed
SAFE
comfort mode (ACC)
Low accelerations
f (distance margin)”
0
importance:
• 0  Ad_D .
• To fix the maximum allowed deceleration value.
• To compute the desired distance in SAFE region.
18/09/2012
N°8
ADAS braking system logic description
IMPOSED ACCELERATION - example
=1
rel distance
rel speed
0
SAFE
0
DESIRED DISTANCE in SAFE region
-1
140
mu=0.2
mu=0.4
mu=0.6
mu=0.8
mu=1
-3
120
-2
-4
-6
-4
-5
-6
100
-8
-7
80
-10
120
60
100
40
80
60
20
40
20
0
0
-8
-9
100
Desired Distance (m)
WARNING
0
DANGER
Relative distance [m]
-2
80
60
40
-10
20
rel distance
rel speed
0
0
10
20
30
40
50
Actual Speed (km/h)
60
18/09/2012
70
80
N°9
Index
1. ADAS braking system logic description (including potential
friction - 0 - infos).
2. Evaluation of 0 knowledge benefits (VIA CarMaker
simulations): critical scenarios’ definition.
3. “ 0 identification algorithm” (longitudinal transient):
experimental tests and results.
4. Implementation of the “ 0 identification algorithm” in
CarMaker & Real-Time analysis.
5. Conclusions.
18/09/2012 N°10
Evaluation of
0
knowledge benefits
μs = μp
= friction
= initial speed V0s = V0p
VS
11
Vehicle P
emergency
brake
VP
≠ friction
μs < μp
= initial speed V0s = V0p
VS
Low Friction Road
μs = μp
= friction
≠ initial speed V0p= 0
VS
Vehicle P
emergency
brake
VP
Normal Road
Vehicle S
emergency
brake in traffic
D Preview Radar
Low Friction Road
18/09/2012 N°11
Evaluation of
0
knowledge benefits
μs = μp
= friction
= initial speed V0s = V0p
VS
12
Vehicle P
emergency
brake
VP
≠ friction
μs < μp
= initial speed V0s = V0p
VS
Low Friction Road
μs = μp
= friction
≠ initial speed V0p= 0
VS
Vehicle P
emergency
brake
VP
Normal Road
Vehicle S
emergency
brake in traffic
D Preview Radar
Low Friction Road
18/09/2012 N°12
Evaluation of
0
knowledge benefits
13
VS
VP
µ=0.4
D1
D2 ≈ 30m
D1
D1 ≈ 15m
ALGORITHM
DESIRED
speed profile
0 knowledge
DOESN’T
IMPROVE
SAFETY
Previous car REAL
speed profile
REAL
speed profile
VP
VS
µ=0.4
D2
D2
DESIRED = REAL
speed profile
Previous car REAL
speed profile
18/09/2012 N°13
Evaluation of
0
knowledge benefits
μs = μp
= friction
= initial speed V0s = V0p
VS
14
Vehicle P
emergency
brake
VP
≠ friction
μs < μp
= initial speed V0s = V0p
VS
Low Friction Road
μs = μp
= friction
≠ initial speed V0p= 0
VS
Vehicle P
emergency
brake
VP
Normal Road
Vehicle S
emergency
brake in traffic
D Preview Radar
Low Friction Road
18/09/2012 N°14
Evaluation of
0
knowledge benefits
15
VS
VP
µ=0.4
D1
µ=1
Speed Profile
D2 ≈ 30m
D1 ≈ 15m
ALGORITHM
DESIRED
speed profile
REAL
speed profile
V
0 knowledge
IMPROVES
SAFETY
µ=0.4
Previous car REAL
speed profile
V
D2
Speed Profile
D3
DESIRED = REAL
speed profile
Previous car REAL
speed profile
18/09/2012 N°15
Evaluation of
0
knowledge benefits
μs = μp
= friction
= initial speed V0s = V0p
VS
17
Vehicle P
emergency
brake
VP
≠ friction
μs < μp
= initial speed V0s = V0p
VS
Low Friction Road
μs = μp
= friction
≠ initial speed V0p= 0
VS
Vehicle P
emergency
brake
VP
Normal Road
Vehicle S
emergency
brake in traffic
D Preview Radar
Low Friction Road
18/09/2012 N°17
Evaluation of
0
knowledge benefits
18
VS
D Preview Radar
µ=0.2
ALGORITHM DESIRED
ACCELERATION
REAL CCELERATION
VS
µ=0.2
D Preview Radar
0 knowledge
IMPROVES
SAFETY
18/09/2012 N°18
Evaluation of
0
knowledge benefits
19
CRASH OCCURS
VS
D Preview Radar
REAL CCELERATION
1. Radar out of range: relative
distance > 60m
Vehicle S
ALGORITHM DESIRED
ACCELERATION
acceleration [m/s 2]
0
Relative Distance [m]
4. DANGER region  algorithm
desired deceleration HIGHER than
real
unknown DESIRED
unknown REAL
-10
23
2. SAFE region (if 0 is unknown)
 no deceleration imposed
3. WARNING region  algorithm
desired deceleration HIGHER than
real
-5
80
24
25
1 2
26
27
Time (s)
3
28
29
4
30
31
UNKNOWN
60
40
20
0
23
24
25
26
27
Time (s)
28
29
30
31
18/09/2012 N°19
Evaluation of
0
knowledge benefits
20
CRASH DOESN’T OCCURS
VS
D Preview Radar
acceleration [m/s 2]
1. Radar out of range: relative
distance > 60m.
Vehicle S
0
Relative Distance [m]
4. DANGER region
unknown DESIRED
unknown REAL
KNOWN
-10
23
2. NO SAFE REGION.
3. WARNING region  as far as
the radar see the vehicle,
deceleration occurs.
-5
24
13
25
26
27
Time (s)
28
4
29
30
31
80
UNKNOWN
60
KNOWN
40
20
0
23
24
25
26
27
Time (s)
28
29
30
31
18/09/2012 N°20
Index
1. ADAS braking system logic description (including potential
friction - 0 - infos).
2. Evaluation of 0 knowledge benefits (VIA CarMaker
simulations): critical scenarios’ definition.
3. “ 0 identification algorithm” (longitudinal transient):
experimental tests and results.
4. Implementation of the “ 0 identification algorithm” in
CarMaker & Real-Time analysis.
5. Conclusions.
18/09/2012 N°21
0
identification algorithm
Higher longitudinal slip
Constant speed
Accelerating/braking
Higher lateral slip
Longitudinal maneuvers
Lateral maneuvers
TARGET: to identify the potential friction before reaching it. As
before as possible!
18/09/2012 N°22
0
identification algorithm
LONGITUDINAL +
TRANSIENT
1st target:
0 MACRO
levels resolution
1.
2.
LOWEST
0 ≤ 0.2
LOW
0.2< 0 ≤0.4
3.
NOT LOW
0.4 < 0
2nd target: to increase resolution with additional levels
1.
2.
LOWEST
0 ≤ 0.2
LOW
0.2< 0 ≤0.4
3.
4.
MEDIUM
0.4< 0 ≤ 0.9
HIGH
0.9 <
0
18/09/2012 N°23
0
identification algorithm
Vizzola Ticino proving ground
Reference test Track
Texture/wet-dry
Reference
T.T.1 (GRB)
Icy-smooth, wet
0.15
T.T.2 (GRA)
Icy-smooth, dry
0.65
T.T.3 (CEA)
Concrete, dry
0.85
T.T.3 (V2A)
Asphalt, dry
1.1
0
1.
LOWEST
0 ≤ 0.2
3.
NOT LOW
0.4 < 0
18/09/2012 N°24
0
identification algorithm
LONGITUDINAL +
TRANSIENT
When 0 is
accessible, over
97% of correct
estimation!
1.
3.
LOWEST
0 ≤ 0.2
NOT LOW
0.4 < 0
18/09/2012 N°25
0
identification algorithm
«…. When 0 is accessible ?...»
LONGITUDINAL +
TRANSIENT
Reference test Track
T.T.1 (GRB)
Icy-smooth, wet
T.T.2 (GRA)
Icy-smooth, dry
T.T.3 (CEA)
Concrete, dry
T.T.3 (V2A)
Asphalt, dry
Texture/wet-dry
Reference
0
0.15
0.25 m/s2
0.65
1.30 m/s2
0.85
1.1
1.50 m/s2
0.75 m/s2
18/09/2012 N°26
identification algorithm
SUBURBAN ROAD
100
[km/h]
80
60
40
20
0
0
200
400
600
800
1000
1200
1400
800
1000
1200
1400
[s]
4
2
[m/s 2]
0
0
-2
-4
0
200
400
600
[s]
18/09/2012 N°27
identification algorithm
Occurrences [%] for each
acceleration class
WETHER CONDITION WHILE TESTS: DRY
20
Occurrencies at each class [%]
0
Highway
Suburban
Urban
15
10
5
0
-3
-2
-1
0
1
2
3
Acc class [m/s 2]
Acceleration classes
(- 3 : 0.1 : +3)
[m/s2 ]:
Cumulative probability [%]
120
Braking (<0)
100
80
60
Accelerating (>0)
40
20
0
-3
-2
-1
0
1
2
3
Acc class [m/s 2]
18/09/2012 N°28
identification algorithm
WETHER CONDITION WHILE TESTS: DRY
Highway
Suburban
Urban
15
0
-3
120
100
80
60
40
-2
-1
20
0
-3
-2
0
1
Acc class [m/s 2]
-1
0
ACCELERATING
5
FREE ROLLING
10
BRAKING
Occurrencies at each class [%]
20
Cumulative probability [%]
0
1
2
3
2
3
Acc class [m/s 2]
[%]
[%]
[%]
Availability
Highway
5.2
89.9
4.9
10.1
Suburban
7.1
88.0
4.9
12.0
Urban
15.5
72.0
12.5
28.0
18/09/2012 N°29
Index
1. ADAS braking system logic description (including potential
friction - 0 - infos).
2. Evaluation of 0 knowledge benefits (VIA CarMaker
simulations): critical scenarios’ definition.
3. “ 0 identification algorithm” (longitudinal transient):
experimental tests and results.
4. Implementation of the “ 0 identification algorithm” in
CarMaker & Real-Time analysis.
5. Conclusions.
18/09/2012 N°30
OFF LINE
Implementation of the
0
identification algorithm
LONGITUDINAL +
TRANSIENT
Emulator for OFF LINE analysis
DATA processing via
simulink code
x, y, speed
Input variable
& imposed 0
REAL TIME
On road test
recording track
and speed profile
DataLogger via AutoBox
dspace
DSpace control desk visualizator
for REAL TIME analysis
CAN state
variables
18/09/2012 N°31
OFF LINE
Implementation of the
0
identification algorithm
LONGITUDINAL +
TRANSIENT
Emulator for OFF LINE analysis
DATA processing via
simulink code
x, y, speed
Input variable
& imposed 0
REAL TIME
On road test
recording track
and speed profile
DataLogger via AutoBox
dspace
DSpace control desk visualizator
for REAL TIME analysis
CAN state
variables
18/09/2012 N°32
Implementation of the
0
identification algorithm
Potential friction level for Left Wheel
Potential friction level for Right Wheel
Vehicle speed
Vehicle acceleration
Example test:
18/09/2012 N°34
Implementation of the
0
identification algorithm
Example test:
18/09/2012 N°35
OFF LINE
Implementation of the
0
identification algorithm
LONGITUDINAL +
TRANSIENT
Emulator for OFF LINE analysis
DATA processing via
simulink code
x, y, speed
Input variable
& imposed 0
REAL TIME
On road test
recording track
and speed profile
DataLogger via AutoBox
dspace
DSpace control desk visualizator
for REAL TIME analysis
CAN state
variables
18/09/2012 N°36
Implementation of the
0
identification algorithm
18/09/2012 N°37
Index
1. ADAS braking system logic description (including potential
friction - 0 - infos).
2. Evaluation of 0 knowledge benefits (VIA CarMaker
simulations): critical scenarios’ definition.
3. “ 0 identification algorithm” (longitudinal transient):
experimental tests and results.
4. Implementation of the “ 0 identification algorithm” in
CarMaker & Real-Time analysis.
5. Conclusions.
18/09/2012 N°38
Conclusions
•
- based, second generation ADAS braking system logic
has been introduced, highlighting efficiency improvement
on driving safety.
0
• Results on 0 identification algorithm (PFI3.2) have been
summarized and method reliability has been estimated via
experimental tests performed on Pirelli proving ground on
reference potential friction tracks.
• Implementation of the 0 identification algorithm (PFI3.2)
in both the CarMaker Simulink for off-line analysis and in
Real Time analysis for data deliverance to beta-version of
new generation ADAS system.
18/09/2012 N°39
Thanks for your attention!
Come to visit us @ the exhibition stand of:
Pezzola & Associati
18/09/2012 N°40