# A Comparative, Experimental Study of Model Suitability to Describe Vehicle Rollover Dynamics for Control Design

Urban and Civil

Nov 16, 2013 (4 years and 5 months ago)

88 views

Dept. Of Mechanical and Nuclear Engineering,
Penn State University

Vehicle Dynamic Modeling for the Prediction
and Prevention of Vehicle Rollover

A Comparative, Experimental Study of Model
Suitability to Describe Vehicle Rollover
Dynamics for Control Design

John T. Cameron

Pennsylvania State University

Dr. Sean Brennan

Pennsylvania State University

Dept. Of Mechanical and Nuclear Engineering,
Penn State University

2
/23

Vehicle Dynamic Modeling for the Prediction
and Prevention of Vehicle Rollover

Outline

1.
Goals

2.
Analytical Vehicle Models

3.
Experimental Model Validation

4.
Conclusions

Dept. Of Mechanical and Nuclear Engineering,
Penn State University

3
/23

Vehicle Dynamic Modeling for the Prediction
and Prevention of Vehicle Rollover

Goals

Examine various vehicle models to determine the
effect that different assumptions have on:

Model order

Model complexity

Number and type of parameters required

Experimentally validate the models to:

Determine model accuracy

Relate modeling accuracy to assumptions made

Determine the simplest model that accurately represents a
vehicles planar and roll dynamics

Dept. Of Mechanical and Nuclear Engineering,
Penn State University

4
/23

Vehicle Dynamic Modeling for the Prediction
and Prevention of Vehicle Rollover

Analytical Vehicle Models

Standard SAE sign convention

Dept. Of Mechanical and Nuclear Engineering,
Penn State University

5
/23

Vehicle Dynamic Modeling for the Prediction
and Prevention of Vehicle Rollover

Analytical Vehicle Models

Basic Assumptions Common to All Models

All models are linear

Result:

Small angles are assumed making cos(
θ
)≈1, sin(
θ
)≈0

Constant longitudinal velocity (along the x
-
axis)

The lateral force acting on a tire is directly proportional to slip
angle

Longitudinal forces ignored

Tire forces symmetric right
-
to
-
left

sin
1
cos

sin
1
cos

tire
tire
C
F

Dept. Of Mechanical and Nuclear Engineering,
Penn State University

6
/23

Vehicle Dynamic Modeling for the Prediction
and Prevention of Vehicle Rollover

Analytical Vehicle Models

Model 1

2DOF Bicycle Model

f
Fu
Kq
q
D
q
M

y
q
f
f
r
r
r
f
f
f
r
f
C
r
V
U
C
l
U
C
U
C
l
U
C
F
F

0
0
0

r
f
r
f
zz
F
F
l
l
y
r
V
mU
r
V
I
m
0
0
2
2
2
2
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0

Dept. Of Mechanical and Nuclear Engineering,
Penn State University

7
/23

Vehicle Dynamic Modeling for the Prediction
and Prevention of Vehicle Rollover

Analytical Vehicle Models

Model 2

3DOF Roll Model

Assumes the existence of a sprung mass

No x
-
z planar symmetry

Originally presented by Mammar et. al., National Institute of Research
on the Transportations and their Security (INRETS), Versailles, France
in 1999

r
f
r
f
s
s
xx
xz
s
xz
zz
s
F
F
l
l
y
gh
m
K
r
V
D
hU
m
mU
r
V
I
I
h
m
I
I
h
m
m
0
0
2
2
2
2
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0

r
f
r
f
zz
F
F
l
l
y
r
V
mU
r
V
I
m
0
0
2
2
2
2
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0

r
f
r
f
zz
F
F
l
l
y
r
V
mU
r
V
I
m
0
0
2
2
2
2
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0

Dept. Of Mechanical and Nuclear Engineering,
Penn State University

8
/23

Vehicle Dynamic Modeling for the Prediction
and Prevention of Vehicle Rollover

Analytical Vehicle Models

Model 3

3DOF Roll Model

Assumes the existence of a sprung mass

x
-
z planar symmetry

Roll
-
steer influence

Originally presented by Kim and Park, Samchok University, South
Korea, 2003

0 0 0 0 0 0 2 2
0 0 0 0 0 0 0 0 2 2
0 0 0 0 0 0
s
f
zz f r
r
s xx s s
m m h V mU V y
F
I r r l l
F
m h I m hU D K m gh
 

  
 
   
 
       
 
 
   
       
    
 
 
   
       
 
 
   
       
  
       
   
 

r
f
r
f
zz
F
F
l
l
y
r
V
mU
r
V
I
m
0
0
2
2
2
2
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0

Dept. Of Mechanical and Nuclear Engineering,
Penn State University

9
/23

Vehicle Dynamic Modeling for the Prediction
and Prevention of Vehicle Rollover

Analytical Vehicle Models

Model 3 (continued)

As a result of the assumption of roll steer, the external forces
acting on the vehicle change accordingly

f
f
r
r
r
f
f
f
r
f
C
r
V
U
C
l
U
C
U
C
l
U
C
F
F

0
0
0
f
f
r
r
r
r
f
f
f
f
r
f
C
r
V
U
C
l
U
C
U
C
l
U
C
F
F

0
*
*
3
,
3
,
Dept. Of Mechanical and Nuclear Engineering,
Penn State University

10
/23

Vehicle Dynamic Modeling for the Prediction
and Prevention of Vehicle Rollover

Analytical Vehicle Models

Model 4

3DOF Roll Model

Assumes a sprung mass suspended upon a massless frame

x
-
z planar symmetry

No roll steer influence

Originally presented by Carlson and Gerdes, Stanford University,
2003

r
f
r
f
xx
zz
F
F
h
h
l
l
y
mgh
K
r
V
D
mU
r
V
I
I
mh
m
2
2
2
2
2
2
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0

r
f
r
f
zz
F
F
l
l
y
r
V
mU
r
V
I
m
0
0
2
2
2
2
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0

Dept. Of Mechanical and Nuclear Engineering,
Penn State University

11
/23

Vehicle Dynamic Modeling for the Prediction
and Prevention of Vehicle Rollover

Analytical Vehicle Models

Effect of assuming force equivalence

Slightly changes plant description (i.e. eigenvalues)

Additionally, causes a higher gain in roll response from the
massless frame assumption

Dept. Of Mechanical and Nuclear Engineering,
Penn State University

12
/23

Vehicle Dynamic Modeling for the Prediction
and Prevention of Vehicle Rollover

Model Fitting Procedures

1.
Experimentally determine the understeer gradient to find the
relationship between front and rear cornering stiffness values.

Considering both frequency and time domains*:

2.
Determine estimates on cornering stiffness values by fitting of
the 2DOF Bicycle Model (Model 1).

3.
Determine estimates on roll stiffness and damping by fitting of
Models 2

4.

*
-

Time domain maneuvers were a lane change and a pseudo
-
step

Dept. Of Mechanical and Nuclear Engineering,
Penn State University

13
/23

Vehicle Dynamic Modeling for the Prediction
and Prevention of Vehicle Rollover

Time Domain Fit Results

10
1
0
5
10
15
Frequency Response, Steering Input to Yaw Rate, Mercury Tracer
U =16.5 Cf =-22750 Cr =-19958.4561 K =38000 D =5000
Mag (dB)
10
1
-100
-50
0
Phase (deg)
Measured
Model 1
Model 2
Model 3
Model 4
10
1
15
20
25
30
35
40
Frequency Response, Steering Input to Lateral Acceleration, Mercury Tracer
U =16.5 Cf =-22750 Cr =-19958.4561 K =38000 D =5000
Mag (dB)
10
1
0
50
100
150
Phase (deg)
Measured
Model 1
Model 2
Model 3
Model 4
Dept. Of Mechanical and Nuclear Engineering,
Penn State University

14
/23

Vehicle Dynamic Modeling for the Prediction
and Prevention of Vehicle Rollover

Model Fitting Results

Results for Steering Input to Lateral Acceleration

10
1
15
20
25
30
35
40
Frequency Response, Steering Input to Lateral Acceleration, Mercury Tracer
U =16.5 Cf =-45500 Cr =-75562.5 K =53000 D =6000
Mag (dB)
10
1
0
50
100
150
Phase (deg)
Measured
Model 1
Model 2
Model 3
Model 4

Freq. Domain Fit

Dept. Of Mechanical and Nuclear Engineering,
Penn State University

15
/23

Vehicle Dynamic Modeling for the Prediction
and Prevention of Vehicle Rollover

Model Fitting Results

Results for Steering Input to Yaw Rate

10
1
0
5
10
15
Frequency Response, Steering Input to Yaw Rate, Mercury Tracer
U =16.5 Cf =-45500 Cr =-75562.5 K =53000 D =6000
Mag (dB)
10
1
-100
-50
0
Phase (deg)
Measured
Model 1
Model 2
Model 3
Model 4

Freq. Domain Fit

Dept. Of Mechanical and Nuclear Engineering,
Penn State University

16
/23

Vehicle Dynamic Modeling for the Prediction
and Prevention of Vehicle Rollover

Model Fitting Results

Results for Steering Input to Roll Rate

10
1
-5
0
5
10
15
Frequency Response, Steering Input to Roll Rate, Mercury Tracer
U =16.5 Cf =-45500 Cr =-75562.5 K =53000 D =6000
Mag (dB)
10
1
-100
-50
0
50
100
Phase (deg)
Measured
Model 2
Model 3
Model 4

Freq. Domain Fit

Dept. Of Mechanical and Nuclear Engineering,
Penn State University

17
/23

Vehicle Dynamic Modeling for the Prediction
and Prevention of Vehicle Rollover

Model Fitting Results

Inconsistency in roll rate measured response does not appear at
lower speeds

Better sensors are required to clarify inconsistencies in data

especially lateral acceleration and roll rate

10
0
10
1
-10
0
10
Frequency Response, Steering Input to Roll Rate
U =8.9 Cf =-45500 Cr =-75560 K =53000 D =6000
Mag (dB)
10
0
10
1
-500
-450
-400
-350
-300
-250
Phase (deg)
measured
Model 2
Model 3
Model 4
Dept. Of Mechanical and Nuclear Engineering,
Penn State University

18
/23

Vehicle Dynamic Modeling for the Prediction
and Prevention of Vehicle Rollover

Remarks on Model Validation

As a result of overall accuracy and simplicity, Model 3
was chosen for further investigation. This entails:

The development of model
-
based predictive algorithms for
rollover propensity

The development of control algorithms for rollover
mitigation

Dept. Of Mechanical and Nuclear Engineering,
Penn State University

19
/23

Vehicle Dynamic Modeling for the Prediction
and Prevention of Vehicle Rollover

Conclusions

A relatively simple dynamic model is capable of
modeling both the planar and roll dynamics of a
vehicle well under constant speed conditions.

Relatively accurate measurements may be taken with
inexpensive sensors

The dynamics are seen even with commercial grade sensors

Important for industry because such sensors are typically
found in production vehicles

Extra care should be taken when model fitting in the
time domain

Dept. Of Mechanical and Nuclear Engineering,
Penn State University

20
/23

Vehicle Dynamic Modeling for the Prediction
and Prevention of Vehicle Rollover

Time Response Tests

Pseudo
-
Step Response, 8.9 m/s, 0.09 rad amplitude, FR Params

1
1.5
2
2.5
0.02
0.04
0.06
0.08
0.1
Step, Steering vs. Time
Time (s)
1
1.5
2
2.5
0
0.05
0.1
0.15
0.2
0.25
0.3
Yaw Rate vs. Time
Time(s)
Measured
Model 1
Model 2
Model 3
Model 4
1
1.5
2
2.5
0
0.2
0.4
0.6
0.8
1
1.2
Lat. Accel. vs. Time
Time (s)
Lat. Accel. (m/s
2
)
2.5
3
3.5
4
0
0.02
0.04
0.06
0.08
0.1
Steering vs. Time
Time (s)
2.5
3
3.5
4
0
0.05
0.1
Roll Rate vs. Time
Time (s)
Dept. Of Mechanical and Nuclear Engineering,
Penn State University

21
/23

Vehicle Dynamic Modeling for the Prediction
and Prevention of Vehicle Rollover

Time Response Tests

Pseudo
-
Step Response, 8.9 m/s, 0.09 rad amplitude, TR Params

1
1.5
2
2.5
0
0.02
0.04
0.06
0.08
0.1
Step, Steering vs. Time
Time (s)
1
1.5
2
2.5
0
0.05
0.1
0.15
0.2
0.25
0.3
Yaw Rate vs. Time
Time(s)
Measured
Model 1
Model 2
Model 3
Model 4
1
1.5
2
2.5
0
0.5
1
Lat. Accel. vs. Time
Time (s)
Lat. Accel. (m/s
2
)

2.5
3
3.5
4
0
0.02
0.04
0.06
0.08
0.1
Steering vs. Time
Time (s)
2.5
3
3.5
4
-0.02
0
0.02
0.04
0.06
0.08
0.1
Roll Rate vs. Time
Time (s)
Dept. Of Mechanical and Nuclear Engineering,
Penn State University

22
/23

Vehicle Dynamic Modeling for the Prediction
and Prevention of Vehicle Rollover

Time Response Tests

Lane Change Maneuver, 17.8 m/s, Right
-
to
-
Left, then Left
-
to
-
Right, FR

0
2
4
6
8
-0.04
-0.02
0
0.02
0.04
Lane Change, Steering Angle vs. Time
Time (s)
0
2
4
6
8
-0.15
-0.1
-0.05
0
0.05
0.1
0.15
Yaw Rate vs. Time
Time(s)
2
4
6
8
-0.5
0
0.5
Lat. Accel. vs. Time
Time (s)
Lat. Accel. (m/s
2
)
Measured
Model 1
Model 2
Model 3
Model 4
0
2
4
6
8
-0.04
-0.02
0
0.02
0.04
Steering Angle vs. Time
Time (s)
0
2
4
6
8
-0.1
-0.05
0
0.05
0.1
Roll Rate vs. Time
Time (s)
Dept. Of Mechanical and Nuclear Engineering,
Penn State University

23
/23

Vehicle Dynamic Modeling for the Prediction
and Prevention of Vehicle Rollover

Time Response Tests

Lane Change Maneuver, 17.8 m/s, Right
-
to
-
Left, then Left
-
to
-
Right, Time

0
2
4
6
8
-0.04
-0.02
0
0.02
0.04
Lane Change, Steering Angle vs. Time
Time (s)
0
2
4
6
8
-0.15
-0.1
-0.05
0
0.05
0.1
0.15
Yaw Rate vs. Time
Time(s)
0
2
4
6
8
-0.5
0
0.5
Lat. Accel. vs. Time
Time (s)
Lat. Accel. (m/s
2
)
Measured
Model 1
Model 2
Model 3
Model 4

0
2
4
6
8
-0.04
-0.02
0
0.02
0.04
Steering Angle vs. Time
Time (s)
0
2
4
6
8
-0.1
-0.05
0
0.05
0.1
Roll Rate vs. Time
Time (s)
Dept. Of Mechanical and Nuclear Engineering,
Penn State University

24
/23

Vehicle Dynamic Modeling for the Prediction
and Prevention of Vehicle Rollover

Experiments Performed

of steering necessary to maintain a steady
-
state turn per g of
lateral acceleration (e.g. units are rad/g)

Provides a relationship between the front and rear cornering
stiffness‘

Lateral acceleration was measured on a 30.5 m radius circle at 6.7,
8.9, and 11.2 m/s

r
f
f
r
us
C
W
C
W
K

2
2
r
f
f
r
us
C
W
C
W
K

2
2
f
us
r
f
f
r
C
K
W
C
W
C

2
Dept. Of Mechanical and Nuclear Engineering,
Penn State University

25
/23

Vehicle Dynamic Modeling for the Prediction
and Prevention of Vehicle Rollover

Model Fitting Procedure

Step 1

Plotting additional steering angle vs. lateral acceleration, the
understeer gradient is simply the slope of the line

y = 0.045x + 0.018
R
2
= 0.9965
0.024
0.026
0.028
0.03
0.032
0.034
0.036
0.125
0.175
0.225
0.275
0.325
0.375
0.425
Lat. Accel (g's)
Dept. Of Mechanical and Nuclear Engineering,
Penn State University

26
/23

Vehicle Dynamic Modeling for the Prediction
and Prevention of Vehicle Rollover

Analytical Vehicle Models

Paper
Model Order
Method of validation
Who are they with
Williams, 1995, Nonlinear control of roll moment distribution…
NL 2DOF
No roll dynamics included, only a "roll moment factor"
Georgia Institute of Technology
Rosam, 1997, Development and simulation of a novel roll…
?
No model or Free Body Diagram Given
University of Bath
Darling, 1998, An Experimental Study of a Prototype…
?
No model or Free Body Diagram Given
University of Bath
Feng, 1998, Automatic Steering Control of Vehicle Lateral...
2 & 3DOF
Errors in published formulation
PATH
Feng, 2000, Decoupling Steering Control For Vehicles…
2 & 3DOF
Errors in published formulation
PATH
Krishnaswami, 1998, A Regularization Approach To Robust…
2DOF
Not enough information given
UMTRI
Wielenga, 1999, A Method for Reducing On Road Rollover…
3DOF
Model formulation not given
Dynomotive
Chen, 1999, A Real Time Rollover ThreatIndex For SUV's
coupled 2DOF
Decoupled approach
UMTRI
Chen, 2001, Differential Braking Based Rollover Prevention…
3DOF
Parameters difficult to obtain
UMTRI
Kitajima, 2000, Control For Integrated Side Slip Roll
8DOF*, 3DOF
Equations complex, not enough information given
UMTRI
Eger, 2003, Modeling of rollover sequences
2DOF
Covers tripped rollovers
University of Karlsruhe, Germany
Kueperkoch, 2003, Novel Stability Control Using SBW…
3DOF
Not relevant to our study
Bosch Corporation
Rossetter, 2003, A Gentle Nudge Towards Safety…
2DOF
Not relevant to our study
Stanford
Takano, 2003, Study on a vehicle dynamics model for…
3DOF
Errors in published information
Tokyo University of Ag. and Tech.
Oh, 2004, The Design of a Controller for the SBW System
9DOF
Model formulation not given
Hyundai/Hanyang University
Paper
Model Order
Who are they with
Sharp, 1993, On the design of an active control system for a…
3DOF
Complex formulation, parameters are difficult to obtain
Cranfield Institute of Technology
Chen, C, 1998, Steering Control of High-Speed Vehicles
2DOF
Not relevant to our study
PATH
Mammar, 1999, Speed Scheduled Vehicle Lateral Control
3DOF
Nicely derived, but no experimental validation. Includes a
Evry University, France
mathematical proof on its model matching abilities.
Cole, 2000, Evaluation Of Design Alternatives For Roll Control…
3DOF
Model is developed through a software package
University of Nottingham
Hyun, 2000, Vehicle Modeling And Prediction Of…
NL 8DOF
Not relevant to our study
Texas A&M
Ikenaga, 2000, Active Suspension Control Of Ground…
7DOF
No description of lateral dynamics
Texas Arlington
Manning, 2000, Coordination Of Chassis Control Systems
NL 5DOF
Not enough information given
University of Leeds, UK
Kim, 2003, Investigation Of Robust Roll Motion Control…
3DOF
Clean presentation, parameters given, model worked
Samchok University, South Korea
Sprague, 2002, Automated stability analysis of a vehicle…
6DOF
Model formulation not given
Exponent Failure Analysis Associates
Huh, 2002, Monitoring System Design For Estimating...
4DOF
No roll dynamics included, only lateral weight transfer
Samchok University, South Korea
Carlson, 2003, Optimal rollover prevention with SBW and diff…
NL4DOF, L3DOF
All work done in simulation
Stanford
Models With Experimental Validation
Models Not Experimentally Validated