A detailed multi-body model for

dynamic simulation of o-road tracked vehicles

D. Rubinstein
a,
*
, R. Hitron
b
a
The Land Systems Unit, Technion Research & Development Foundation, Haifa 32000, Israel
b
Faculty of Agricultural Engineering, Technion Israel Institute of Technology, Haifa 32000, Israel
Available online 23 April 2004
Abstract
Currently available models for dynamic simulation of tracked vehicles usually include su-
per-elements to describe the tracks and the suspension systems. In these models, the dynamics
of the track, the interaction between each track link and the ground, and their eect on the
vehicle dynamics cannot be considered properly. The rapid increase in computing speed en-
ables the utilization of more complex models, including numerous bodies and force elements.
A three-dimensional multi-body simulation model for simulating the dynamic behavior of
tracked o-road vehicles was developed using the LMS-DADS simulation program. The
model incorporates detailed description of the track, the suspension system, and the dynamic
interaction between its components. The bodies of the model are the chassis, the wheel-arms,
the wheels, and each track link. Three-dimensional contact force elements are used to describe
the interaction of the track links with the vehicles road wheels, sprocket, and idler. Additional
force elements are used to simulate the bump stops and the dampers. User-dened force el-
ements are used to describe the interaction between each track link and the ground. The
normal and tangential forces are calculated using classical soil mechanics equations, such as
Bekker and Janosi correlations. Sinkage and slip are calculated separately for each track link.
Alternative correlations, based on recent studies of the dynamic variations of these forces, can
also be used. The model was rst applied to the M113 armored carrier. Simulation results
under various road conditions were compared with the results of a super-element-based
model. It was concluded that the inuence of the track dynamics and the soillink interaction
on the vehicle dynamics can be better predicted with the newly developed model.
2004 ISTVS. Published by Elsevier Ltd. All rights reserved.
*
Corresponding author. Fax: +972-4-8224580.
E-mail address: agrdror@tx.technion.ac.il (D. Rubinstein).
0022-4898/$20.00 2004 ISTVS. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.jterra.2004.02.004
www.elsevier.com/locate/jterra
Journal
of
Terramechanics
Journal of Terramechanics 41 (2004) 163173
Keywords: Multi-body simulation; Tracked vehicle; Track link; Track-spud
1. Introduction
The suspension unit of a tracked vehicle consists of many components, which need
to be optimized for proper design. Optimization based on eld tests may signicantly
increase development cost and time. Modern design of a tracked vehicle, based on a
simulation program that represents the suspension units, enables analysis of the ve-
hicles performance prior to its nal design and eld-testing. During the last decade,
many such simulation programs have been developed and adapted for this purpose.
However, the physical models, especially for tracked vehicles, still need improvement
[1,2]. Improving the simulation model may reduce the number of eld tests required.
Vehicle simulation programs are divided into two categories: the rst consists of
special-purpose codes and the second is composed of multi-body programs [3]. As
computers have become increasingly powerful, the multi-body programs have be-
come a popular means for simulating the dynamic behavior of vehicles [4].
Available 2-D special-purpose programs for o-road vehicles, such as VEHDYN
I and II [5], are time-ecient, but include over-simplied assumptions (such as
constant forward velocity), and are not designed for detailed analysis of the sus-
pension system. The multi-body codes are much more powerful programs. However,
most of them are oriented to road-vehicle simulation [4,6], and are not suitable for
tracked-vehicle design. The DADS and ADAMS programs are currently the most
widely used multibody codes [6].
Burt [7] has reviewed some of the soilwheel and soilwheeltrack interaction
models. The present survey includes some of the empirical and semi-empirical soil
wheel models, as well as models basedonconstitutive relations. The radial springmodel
is widely used in simulation programs to represent the soilwheel interaction [1,5,8,9].
The purpose of the study is to present a detailed multi-body model for dynamic
simulation of o-road tracked vehicles.
2. Mathematical modeling
2.1. Single track-link and soil interaction model
The ground level is described using the XYZ world-coordinate system. The ground
level may be depicted in any prole as the following function.
Z f X; Y : 1
The track link is spatially oriented, with coordinate system x
0
y
0
z
0
attached to the link.
Where y
0
is in the longitudinal direction of the link, x
0
is the lateral direction and z
0
is
perpendicular to x
0
and y
0
. Another coordinate system X
0
Y
0
Z
0
is parallel to x
0
y
0
z
0
and
its origin is intercepted by XYZ origin. The link may penetrate into the soil when the
sinkage is measured along the z
0
-axis. A two-dimensional plot of a ground prole
164 D. Rubinstein, R. Hitron / Journal of Terramechanics 41 (2004) 163173
with a sinking link is presented in Fig. 1. A view of the link and its coordinate
systems from above is shown in Fig. 2.
In these gures:
^u
1
; ^u
2
; ^u
3
unit vectors of the X; Y ; Z axes directions, respectively
^u
0
1
; ^u
0
2
; ^u
0
3
unit vectors of the X
0
; Y
0
; Z
0
axes directions, respectively
origin of the x
0
y
0
z
0
coordinate system
g point at which the z
0
-axis passes through ground-surface level
x

; y

; z

location of point in the XYZ coordinate system

x
0

; y
0

; z
0

location of point in the X

0
Y
0
Z
0
coordinate system
x
g
; y
g
; z
g
location of point g in the XYZ coordinate system
x
0
g
; y
0
g
; z
0
g
location of point g in X
0
Y
0
Z
0
coordinate system
a link length
b track width
Dsinkage of the track link in z
0
-direction.
Fig. 2. View of the link and its coordinate systems from above.
Fig. 1. Two-dimensional description of a track-link sinkage.
D. Rubinstein, R. Hitron / Journal of Terramechanics 41 (2004) 163173 165
An unique denition of X
0
Y
0
Z
0
orientation (the same as the track link) may be
obtained using the transformation matrix A, where
A
a
11
a
12
a
13
a
21
a
22
a
23
a
31
a
32
a
33
_
_
_
_
and a
ij
^u
i
^u
0
j
i; j 1; 2; 3: 2
Theoretically, the values of matrix A components may be calculated based on three
independent parameters. In our case, we used four dependent Euler parameters
(e
0
; e
1
; e
2
; e
3
) as follows:
a
11
e
2
0
e
2
1
e
2
2
e
2
3
;
a
22
e
2
0
e
2
1
e
2
2
e
2
3
;
a
33
e
2
0
e
2
1
e
2
2
e
2
3
;
a
ij
2 e
i
e
j
_
e
0
e
k
_
i; j; k 1; 2; 3; i 6 j; k 6 i; k 6 j:
3
The dependency equation of the Euler parameters is
e
2
0
e
2
1
e
2
2
e
2
3
1: 4
The unit vectors of the relative coordinate system X
0
Y
0
Z
0
is obtained based on
world coordinate system directions and Euler parameters as follows:
^u
0
i

3
j1
a
ji
^u
j
i 1; 2; 3: 5
The world coordinates x

; y

and z

, as well as the Euler parameters, are gen-

eralized coordinates and their values are known in each time step of the simulation.
Therefore, point location is calculated as a function of the known parameters.
x
0

^u
1
y

^u
2
z

^u
3
_ _
^u
0
1
;
y
0

^u
1
y

^u
2
z

^u
3
_ _
^u
0
2
;
z
0

^u
1
y

^u
2
z

^u
3
_ _
^u
0
3
:
6
The X
0
and Y
0
coordinates of the point g are identical to the coordinate of the x
0
y
0
z
0
origin and the Z
0
coordinate is z
g
. Thus, the world coordinates of g are:
x
g
x
0

^u
0
1
y
0

^u
0
2
z
0
g
^u
0
3
_ _
^u
1
;
y
g
x
0

^u
0
1
y
0

^u
0
2
z
0
g
^u
0
3
_ _
^u
2
;
z
g
x
0

^u
0
1
y
0

^u
0
2
z
0
g
^u
0
3
_ _
^u
3
:
7
Substitution of Eq. (7) into Eq. (1) yields a non-linear algebraic equation where
the unknown is z
0
g
. The equation is solved simply and eciently using iterations. We
found that in most of the cases no more than four iterations were required for the
solution. The sinkage distance along the z
0
coordinate is
166 D. Rubinstein, R. Hitron / Journal of Terramechanics 41 (2004) 163173
D z
0
g
z
0

: 8
When the track link is in contact or sinking into the soil, the value of D is positive.
Otherwise, when the link is above the soil, the value is negative. The velocity v

of the
point on the track link is given in the world coordinate system as follows:
v

_ x

^u
1
_ y

^u
2
_ z

^u
3
: 9
The sinkage velocity
_
D is the projection of the link velocity on the negative
direction of the z
0
-axis.
_
D v

^u
0
3
: 10
The track-soil model proposed by Bekker [10] is based on the stressdisplacement
relationship for a single application of load to the soil. The original formula has been
modied by adding a viscous friction element. The normal direction of the pressure
sinkage and sinkagevelocity relationship is
p
k
c
b
k
u
_ _
D
n
C
_
D for D > 0 and p > 0;
0 otherwise;
_
11
where p is the pressure, b the track width, C the damping per unit area coecient and
k
c
; k
/
; n are the empirically determined constants.
Note that Bekker [10] originally proposed a vertical sinkage. In our case, the
sinkage is perpendicular to the track link plane surface. When the link penetrates
into the soil, the forces of the two components of Eq. (11) are in the same di-
rection. During the rebound, the forces are in opposite directions. Therefore, a
value of zero pressure may be obtained while geometrically the link is still sinking
into the soil. This is the point at which the link and the soil actually part from
one another.
The attainable locomotion of the whole vehicle over a terrain is based on the shear
forces that develop between track links and soil in the longitudinal directions of the
links. The shear stressdisplacement relation is obtained as follows:
s c p tan / 1
_
e

r j j
k
_
; 12
where s is the shear stress, D the shear displacement, c the cohesion, u the angle of
internal friction, k the empirically determined constant.
The value of the shear displacement is
r
_
v

^u
0
2
dt: 13
Integration of Eq. (13) is not activated unless a contact between the link and the soil
is detected. The integration starts from the beginning with each contact. Assuming
constant distributed shear stress yields the shear force F
s
as follows:
F
s
sign v

^u
0
2
_ _
sab; 14
D. Rubinstein, R. Hitron / Journal of Terramechanics 41 (2004) 163173 167
where
signx
1 x > 0;
1 x < 0;
0 x 0:
_
_
_
15
For side-slide modeling, we assume a model for lateral direction that is similar to
the longitudinal direction. Thus:
F
L
s
sign v

^u
0
1
_ _
ab c p tan / 1
_
e

r
L
j j
k
L
_
16
and
r
L

_
v

^u
0
1
dt: 17
The model can be rened by dividing the link area into n sub-areas (n 4 is
reasonable) and applying the model described above to each sub-area. This modi-
cation has been checked using a single-link simulation model without any signi-
cant eect. However, it has not been checked on the whole vehicle model.
2.2. Vehicle model
The M113 armored carrier vehicle was selected as a typical model of tracked
vehicle. The vehicle consists of a main body (chassis) and two track systems. Each
track system contains a track with 63 links, 5 road-wheels and road-arms, a sprocket,
and an idler. The road-arms of the M113 are relatively parallel. Each road-arm is
attached by a torsional rod to the chassis, and there is a translational damper be-
tween the road-arm and the chassis. While riding over an obstacle, the road-arm/
wheel may receive an impact. The impact moves the wheel toward the chassis. The
suspension components (torsional rod and damper) restrain the relative motion of
the road-arm and reduce the force exerted onto the chassis. When the force of the
impact is very high, the relative motion is halted by the bumper. A description of
these components is provided in Fig. 3.
The LMS-DADS multi-body simulation program is used for the vehicle model-
ing. The soillink interaction model (described in the previous section) was added to
the program as a user-force subroutine. The links are connected to each other using
revolute joints. In order to prevent excessive constraints, the joint between the last
link and the rst link is dened by preventing the relative motion in y
0
- and z
0
-
directions. Additional planer joints between one link and the sprocket or idler is
required for keeping the track in the sprocket-idler plane. The friction in each axis
connected between links is considered by adding force elements (RSDA) to the
joints. The contact of the links with the road-wheel, sprocket, and idler is formulated
with a contact element as a non-linear spring and damper or Hertz theory. The
friction between the contact surfaces is also considered. The connections of the road-
wheels with the road-arms and the road-arms with chassis are implemented with
revolute joints. Force elements with non-linear characteristics (TSDA and RSDA)
168 D. Rubinstein, R. Hitron / Journal of Terramechanics 41 (2004) 163173
are used for the modeling of the torsional bars and the dampers. The implementation
of the bump-stops can be created based on the available components of the contact
elements. A user can add its own model when a more comprehensive one is required
(for example, for bump-stops with hydropneumatic components).
2.3. Power train
The engine torque is applied on the sprocket through the power train system.
Every engine type has a characteristic curve with a specied working zone. The curve
is a torque or power vs. rotating speed. Note that the torque value should be dened
by the riding conditions and not by the characteristic curve. The characteristic
species only the maximum available torque or power in a certain condition. The
rotating mass of inertia of the engine and the power train components is relatively
small (compared to those of the chassis and track links), and can be neglected
without any signicant eect on the whole vehicles dynamic behavior. Therefore, the
use of a detailed and accurate model for the power train system is not important
unless the simulation goal is to explore the behavior of the power train components.
The engine and the power train system are considered part of the main body. The
equivalent rotating mass of inertia can be added to that of the sprocket.
In the simulation model, the engine torque is applied directly on the sprocket.
Therefore, the torque of the characteristic curve should be multiplied by the trans-
mission ratio and the rotation speed divided by the transmission ratio. Each gear has
a dierent transmission ratio. Thus, instead of one characteristic curve, one can
obtain as many curves as the number of the gears. We propose a propulsion model as
described in the block diagram (Fig. 4).
The reference velocity of the vehicle v
ref
can be constant or a function of time,
location, and riding conditions. Subtracting actual velocity v from the reference
velocity yields the error signal. The required torque T
r
is obtained by applying the
cruise controller on the error signal. The controller can be proportional where the
Fig. 3. Description of a track system.
D. Rubinstein, R. Hitron / Journal of Terramechanics 41 (2004) 163173 169
gain is the driver behavior. The required torque signal goes through the character-
istic limitation box. The signal is checked; if it is above the characteristic curve, its
valve is changed to the maximum accessible. The output of characteristic limitation
is the actual torque T
a
that is applied on the sprocket. The current vehicle velocity v
and the sprocket rotational speed x are the output of the DADS plant. The rota-
tional speed is used for proper gear selection and denition of the current torque
limit. The velocity returns to the beginning of the control loop.
3. Case study
The M113 armored carrier vehicle was selected for examination of the simulation
results. The case study focuses on single track-link behavior, a comparison with
Wong [11], and a comparison of the current track model with the DADS super-
element model.
3.1. Single track link
The single link was checked by simulation of shearing and sinkage tests. In the
shearing simulation, the link was constrained with uniaxial quasistatic motion with
three levels of applying loads. The simulation results are shown in Fig. 5. In the
gure, the shear stress is normalized with the normal stress.
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
shear displacement [m]
n
o
r
m
a
l
i
z
e
d

s
h
e
a
r

s
t
r
e
s
s

Vertical load - 1 kN

Vertical load - 10 kN
Vertical load - 5 kN
Fig. 5. Normalized shear stress vs. shear displacement.
Fig. 4. Block diagram of the propulsion model.
170 D. Rubinstein, R. Hitron / Journal of Terramechanics 41 (2004) 163173
The cohesion aects the results when a relatively small normal stress is applied.
When the normal stress is increased, the internal friction inuences the results and
the cohesion is neglected. When the vertical load is 1 kN, the maximum shear stress
is 0.68 of the normal stress. As the load increases, the maximum stress approaches a
value of 0.6, which is the internal friction value.
A simulation of a sinkage test was performed, loading with constant sinking
velocity and unloading with the same velocity in the opposite direction. The soil
damping per unit area coecient is 500,000 Ns/m
3
. The sinkage results are provided
in Fig. 6.
Two sinking velocities are presented in Fig. 6 1 and 20 cm/s. The damping eect
is expressed in the dierence between the loading and unloading curve. In the higher
velocity, the dierence between the curves is about 200 kPa and in the lower velocity,
the curves are almost identical.
3.2. Normal ground pressure
Wong [11] published test results of the normal pressure distributions under the
M113 track in various types of terrain and slip levels. The test results were
compared with simulation results of NTVPM-85. The test and the model
(NTVPM-85) were compared with the presented simulation model on lete sand
with a slip level of 6.1%. The model for the comparison was built to resemble the
vehicle used in Wongs tests as closely as possible. Only part of the necessary
information about the vehicle and the terrain conditions was found. Therefore,
the rest of the information was based on other sources. However, this data is not
necessarily correct. In Wongs tests, a drawbar force cased the slip. In our
model, a drawbar force was applied at the vehicles center of gravity in a
horizontal direction. Note that the point where the force is applied and its di-
rection may signicantly aect results. The comparison results are shown in
Fig. 7.
The graphs in Fig. 7 are characterized by a sequence of ve peaks. Each peak is
related to the pressure under a road-wheel. These results are sensitive to variation in
soil properties, the location of the drawbar force, and its direction. Therefore, as
Fig. 6. Loading and unloading track link in constant sinking velocity.
D. Rubinstein, R. Hitron / Journal of Terramechanics 41 (2004) 163173 171
explained above, only qualitative agreement between the test results and the model
can be achieved.
3.3. Riding simulation
Riding simulation over a ditch (5 m long and 20 cm depth) was performed with
5 m/s forward velocity. The results of the new model were compared with those of
the existing DADS model. The FFT transformations of the simulation results are
presented in Fig. 8. The results of the two models are fairly similar until 20 Hz.
Higher frequency results in signicant dierences between the models. This is
probably due to the dynamic behavior of the track link, which is eectively taken
into account in the new model.
0.00
0.01
0.10
1.00
10.00
0 20 40 60 80 100
frequency [Hz]
A
c
c
e
l
e
r
a
t
i
o
n

[
m
/
s
2
]
NEW MODEL
DADS MODEL
Fig. 8. FFT Transformation of the mass center vertical acceleration riding over a ditch with forward
velocity 5 m/s.
0
50
100
150
200
250
300
350
400
450
0 1 2 3
Distance [m]
P
r
e
s
s
u
r
e

[
k
P
a
]
4
Wong Measured Wong Predicted
New model
Fig. 7. Comparison between measured and predicted ground pressure.
172 D. Rubinstein, R. Hitron / Journal of Terramechanics 41 (2004) 163173
4. Conclusion
Accurate model for track link and terrain interaction based on Bekker and Ja-
nosis approaches was developed. A detailed model of a tracked-vehicle suspension
was built using the LMS-DADS simulation program. The link model was included
in the suspension model.
The single-link simulations show that the link model operates reasonably well.
The normal stress under the track was predicted using the model and compared with
test results [11]. A good qualitative matching between the predicted stress and the
measured stress was obtained.
Comparison of the new model with the DADS model in riding over an obstacle
was performed. The comparison results indicate a good match at lower frequency
and a signicant dierence at higher frequency.
References
[1] Rubinstein D, Galili N. REKEM a design-oriented simulation program for o-road track vehicles. J
Terramech 1994;31:32952.
[2] Rubinstein D. An upgraded o-road track model for multibody simulation. In: Proceeding of the
Eighth European Conference on ISTVS, Umeo, Sweden, 2000.
[3] Kortum W, Sharp RS. Multibody computer codes in vehicle system dynamics. Amsterdam: Swets &
Zeitlinger; 1993.
[4] Schiehlen W. Multibody system handbook. Berlin: Springer-Verlag; 1990.
[5] Creighton DC. Revised vehicle dynamics module: Users guide for computer program VEHDYN II,
Technical Report No. SL-86-9, VSAE-WES, Vicksburg, MS, 1986.
[6] Sharp RS. The application of multi-body computer codes to road vehicle dynamics modeling
problems. J Automobile Eng 1994:5561.
[7] Burt EC. Soil-tire/track interaction-current and future research needs. J Terramech 1993;30:31723.
[8] Shneor Y. Dynamic modeling of o-road tracked vehicle suspension system, MSc Thesis, Technion
Israel Institute of Technology, 1988.
[9] McCullough MK, Haug EJ. Dynamics of high mobility track vehicles. J Mech, Trans Automat Des
1986;108:18997.
[10] Bekker MG. O-the-road locomotion. Ann Arbor: The University of Michigan Press; 1960.
[11] Wong JY. Terramechanics and o-road vehicles. New York: Wiley-Interscience; 1989.
D. Rubinstein, R. Hitron / Journal of Terramechanics 41 (2004) 163173 173