See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/308968871

Full scale truck steering-system modeling and validation

Conference Paper · September 2016

CITATIONS READS
0 182

3 authors:

Jan Loof Igo Besselink

Technische Universiteit Eindhoven Technische Universiteit Eindhoven
9 PUBLICATIONS   5 CITATIONS    57 PUBLICATIONS   729 CITATIONS   

SEE PROFILE SEE PROFILE

Henk Nijmeijer
Technische Universiteit Eindhoven
886 PUBLICATIONS   18,423 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Cooperative automated maneuvering View project

Truck merging support View project

All content following this page was uploaded by Jan Loof on 10 October 2016.

The user has requested enhancement of the downloaded file.

Full scale truck steering-system modeling and validation

Jan Loof, Igo Besselink and Henk Nijmeijer

Department of Mechanical Engineering, Dynamics and Control Group
Eindhoven University of Technology, the Netherlands

ABSTRACT: In this paper a multi-body 44-DOF tractor semi-trailer model is coupled to a 4-DOF steering-
system which includes friction and hydraulic power-steering. An extended wheel hub geometry is used to pro-
vide the correct feedback torque from the wheels. A tie-rod has been included to include Ackerman steering. An
instrumented tractor semi-trailer is used to verify the steering-model predictions during driving. The focus lies
on the prediction of the steering-wheel torque when the vehicle velocity and steering-wheel angle is given as
an input for the simulation. Two tests are discussed in this paper, a J-turn at 80 km/h and a sinusoidal steering-
wheel input with a frequency of 0.4 Hz at 65 km/h. The comparison of the measured signals and the predicted
values shows an accurate prediction and of the steering-system model. The non-linearities caused by the friction
and the hydraulic assistance system can clearly be seen in both the measurement and the simulation.

1 INTRODUCTION ulating the steering angle of the front wheels in trucks

The amount of vehicles on the road worldwide is
increasing every day. In 2010 more than 1 billion
vehicles were registered worldwide. In 2016 a total
35 million heavy-duty trucks are registered. The
amount of trucks is also expected to grow in the
future, by 2030 a growth of 17 percent is expected
in Europe while in Asia and the Middle East this
growth is expected to be more in the direction of 150
to 200 percent [7]. Statistics show that trucks and in
particular tractor semi-trailers are over-represented in
accidents in comparison to cars. Several works show
that causes for this lie in poor rear and side visibility
leading to blind spot crashes [10, 12]. Since truck
drivers are often exposed to long working times with
a low concentration level this can result in fatigue
and also in high-way hypnosis which can lead to
loss-of-control accidents and road departures [11].
Furthermore the size of trucks is growing as well,
this also involves extra articulations which result
in complexer dynamics [2]. To support the driver
under difficult circumstances a number of support
systems are currently available on high-end trucks or
are being developed. These systems include but are
not limited to: anti-lock braking system, electronic
stability control, adaptive cruise-control, intelligent
speed adaption, forward collision warning, automatic
emergency braking, blind spot information, lane
departure warning and curve-speed warning.
At the moment of writing, lateral control by manip-
is not applied in a lot of vehicles on the market. As
a part of a Dutch STW project called ’Truck Merg-
ing Support’, a merging and lane changing/keeping
assistant is being developed to improve safety on the
roads. To investigate the possibilities for supported
truck driving the current steering-system is first mod-
eled and analyzed. Literature in passenger cars reveals
that the modeling of the hydraulic power-steering sys-
tem is necessary [3, 9] and that friction cannot be ne-
glected [8] especially around the center position. Pre-
vious works [5, 6] describe the modeling and model
verification of a steering-system of a commercial ve-
hicle by means of test-bench measurements outside
the vehicle. The aim in these works was to predict in-
put and output angles, input and output torques and
pressures in the steering-system. This has resulted in
the model shown in Figure 4.
In this paper the previously modeled steering-
system model is implemented in a multi-body truck
model. A number of tests is performed with a real in-
strumented vehicle on a proving ground. The model
predictions are compared with the measured signals to
verify the models’ validity. This paper is organized as
follows: In section 2 the vehicle model and steering-
system model will be discussed, section 3 discusses
the instrumentation of the test vehicle, section 4 dis-
cusses the performed tests and the model prediction,
section 5 contains the conclusions and recommenda-
tions, section 6 contains the acknowledgments and
section 7 contains the future work.
2 VEHICLE AND STEERING-SYSTEM MODEL
As a vehicle model a 44 degree of freedom (DOF)
multi-body truck-semitrailer model is used. As shown
in Figure 1, this model contains five axles in total,
the chassis of the tractor and trailer contain torsional
flexibility and an advanced cabin model is used. This
model is constructed by making use of MATLAB
SimMechanics. The tyre-road interface is taken care
of by usage of TNO-Delft tyre.
Figure 1: 44 DOF multi-body truck model as in [4] (the numbers
in the figure indicate the degrees of freedom for that particular
mass).
A steering-system of a general truck contains com-
ponents as in Figure 2. The driver controls the
steering-wheel which is connected to the steering-
house via the steering-column and two universal
joints. These universal joints are necessary to facil-
itate the height adjustment and for packaging rea-
sons. The steering-house contains a hydraulic power-
steering system which amplifies the input torque gen-
erated by the driver. The output of the steering-house
is connected to the pitman-arm which moves the drag-
link. This drag-link movement is converted into a
steering motion of the left wheel by means of the
wheel lever. The left wheel is connected via a tie-rod
to the right wheel.
Figure 2: Components in a typical truck steering-system [1]
The steering-system model (Figure 4) is based on
[5, 6]. The steering-wheel is model by an inertia,
Jsw , with friction, Tf ric,sw , due to the bearings. The
driver supplies the steering-wheel torque Tsw and
the steering-wheel has angle δsw . The steering-wheel
is connected to the input of the steering-house via
two universal joints. All the stiffness in between the
steering-wheel and the steering-house is combined in
Top view (toe-in)
x
y
δ0
Left side view Rear view
Steering-axis
τ γ0 σ
Steering-axis
Wheel centre King-pin
centre
z0
z
z
x
y
rτ,0 rσ,0
Figure 3: Left wheel hub geometry with caster angle τ , king-pin
inclination angle σ, static camber angle γ0 , static toe angle δ0 ,
static caster offset rτ,0 , static wheel center offset rσ,0 and static
vertical hub offset z0 .
one spring ksc . The input of the steering-house has
angle δsh,in and inertia Jsh,in . Due to the bearings
there friction, Tf ric,hua , is assumed here as well. The
input of the steering-house is connected to the output
via two springs in series which represent the torsion-
bar and the spindle, respectively ktb and ksp . A small
damper dhua is added to include material damping. A
fixed gear ratio ish is used for the gear ratio caused
by the spindle and sector shaft. The output of the
steering-house has inertia Jsh,out and angle δsh,out .
Friction Tf ric,ha is present here as well due to the seals
of the hydraulic piston. The power-steering system
is modeled as a symmetric Wheatstone bridge with
the deflection of the torsion-bar as an input and the
pressure difference across the cylinder as an output.
All the stiffness in between the steering-house and
the left wheel is captured in one parameter kha . The
kinematic ratio caused by the pitman-arm, drag-link
and wheel-lever is drawn as a gear-ratio, in the model
it is implemented by making use of the kinematics.
The left wheel has inertia Jkp,L and angle δkp,L , the
friction caused by the king-pin bearings is modeled
as Tf ric,kp,L . The right wheel has inertia Jkp,R and
angle δkp,R , again the friction is modeled by Tf ric,kp,R .
To generate the correct feedback torque from the
wheels to the steering-system the wheel hub geom-
etry contains the caster angle, king-pin inclination
angle, static camber angle, static toe angle, static
caster offset, static wheel center offset and a static
vertical hub offset as shown in Figure 3 for the left
wheel. Both wheels are connected via the tierod with
stiffness ktierod and include Ackerman steering.
 Steering-house
Steering-wheel Wheatstone bridge
Tf ric;sw Tf ric;hua
+ -
Tsw U-joint 1 ksc ktb
U-joint 2
Jsw , δsw
∆tb
Jsh;in , δsh;in dhua
Figure 4: Steering-system model used in the simulations, this model is based on [5, 6]
3 TEST VEHICLE
To verify the steering-system model in combination
with the multi-body truck model an instrumented
tractor semi-trailer is used. The vehicle is equipped
with various sensors, the sensors which are used in
this paper are highlighted in Figure 6. The cabin is
equipped with a 3-axis accelerometer (axC , ayC , azC )
and a 3-axis gyroscope (rxC , ryC , rzC ) at the driver
position. The front and rear part of the chassis are
equipped with a lateral accelerometer (ayF and ayR )
and a gyroscope which measures the yaw-rate (rzF
and rzR ).
The steering-system is instrumented as shown in
Figure 5. The steering-wheel uses a measurement
steering-wheel to measure the steering-wheel torque
and steering-wheel angle (δsw and Tsw ). The pitman-
arm angle, δpa , is measured as well as both king-pin
angles (δkp,L and δkp,R ). To measure the force trans-
ferred between the wheels and the steering-house, the
drag-link force, Fdl is measured.
4 VERIFICATION OF THE MODEL
A number of tests are executed with the instrumented
vehicle on a proving ground. The following tests have
been done:
Figure 5: Overview of the sensors in the steering-system, here
the following nomenclature is used: sw = steering-wheel, pa =
pitman-arm, kp = king-pin, dl = drag-link, L = left, R = right, δ
= angle, T = torque and F = force.
Left wheel Right wheel
Tf ric;kp;L Tf ric;kp;R
TP S Tf ric;ha
kha
ksp
ipa Jkp;L ktierod Jkp;R
ish δkp;L δkp;R
Jsh;out , δpa
• Steady-state cornering between standstill and 65
km/h.
• J-turns at 80 km/h.
• A sinusoidal input at 65 km/h, frequency content
between 0.1 and 2.7 Hz.
All the tests have been performed by a driver
and not with a steering-robot, therefor there can be
some difference between similar tests and sinusoidal
inputs may not always be a 100 % symmetric. The
steady-state cornering tests have been used to tune
the cornering stiffness and the self aligning stiffness
of the tyres. The J-turns are used to evaluate the
model response to a step input and the sinusoidal
input tests can be used to assess the model accuracy
as a function of frequency. Due to the limited space
in this paper two experiments will be discussed; a
J-turn since it contains information about on-center
driving, a dynamic response and the steady-state
behavior. Also a sinusoidal input with a frequency
of 0.4 Hz will be discussed since this is a frequency
which occurs a lot during high-way driving. Due
to confidentiality all graphs in this paper have been
scaled and made unit-less.
In order to simulate the experiment it is important
that the virtual vehicle has the same forward velocity
as the real vehicle. To do so a cruise-controller is
implemented in the model to control the forward
velocity. The measured forward velocity is used as
an input for the model. Furthermore the measured
steering-wheel angle is used as an input for the
model. As a road input for the model a flat road is
assumed. In reality the road is never exactly flat and
some inclination is always present. The assumption
is made here that the road roughness can be neglected
as well as the inclination.
Figure 7 shows the measured and predicted signals
in the steering-system during this J-turn. A left-hand
turn is made by applying a step-steer input after 2
seconds. The vehicle speed during this experiment is
kept constant by means of the cruise control system.
 Cabin 3 axis accelerometer (axC , ayC , azC ) z
Cabin 3 axis gyroscope (rxC , ryC , rzC ) y x

Chassis lateral acceleration front and rear (ayF , ayR )

Chassis yaw-rate front and rear (rzF , rzR )

Figure 6: Side view of the motion sensors in the truck, here the following nomenclature is used: C = cabin, F = front, L = left, R =
rear, a = accelerometer and r = gyroscope

The final steering-wheel position is held until the end Data

Prediction 2
of the test. The angles are predicted accurately by the 10

Tsw [−]
δsw [−]
model, the king-pin angles show a small offset which 1
seems to be a bit larger right than left. The drag-link 0
force is predicted accurately as well, during straight- 0
-10
line driving a small difference is seen, probably due 0 5 0 5
to the road inclination which is not present in the
model. The step-response and end value appear to 0.3 0.3

δkp,R [−]
δkp,L [−]

be predicted well. The steering-wheel torque already 0.2 0.2

shows a lot of dynamics even when driving straight. 0.1 0.1
Since there is friction present in the system and the 0 0
steering-wheel inertia is also taken into account small 0 5 0 5
movements of the steering-wheel will be registered
as well. It is important to predict this behavior as 0.8
0.3
well since in contributes to the on-center feeling. Fdl [−] 0.6
δpa [−]

0.2 0.4
The peak value at 2 seconds is caused by the sudden 0.1 0.2
change and is largely steering-wheel inertia related. 0 0
The steady-state value is mainly caused by the tyre 0 5 0 5
feed-back and the power-steering system. Figure 8 Time [s] Time [s]
shows the accelerometers and gyroscopes. Interesting
to see here is that the front and rear accelerometer Figure 7: Steering-system signals during a J-turn
signal differ quite a lot, especially when the step input
is applied. The model predicts similar behavior as 1 Data 1
well. The yaw-rates front and rear appear to be quite Prediction
ay,F [−]

ay,R [−]

0.5 0.5
similar. To evaluate what the driver experiences the
lateral acceleration in the cabin and the roll-rate in 0 0
the cabin is evaluated. Both are predicted accurately -0.5 -0.5
as well. Interesting to see here is that the lateral 0 5 0 5
acceleration in the cabin differs from both chassis
1 1
measured lateral accelerations again when the step is
rz,F [−]

rz,R [−]

applied. 0.5 0.5

0 0
Secondly, a sinusoidal input with a frequency of 0 5 0 5
0.4 Hz at a vehicle speed of 65 km/h is evaluated.
Also during this experiment the vehicle speed is 1.5
0.4
1
ay,C [−]

rx,C [−]

kept constant by the cruise-control system. Figure 9

0.2 0.5
shows the steering-system signals. Again the angles 0
in the steering-system are predicted with a high 0 -0.5
accuracy. The drag-link force shows some small 0 5 0 5
mismatches around the peaks for negative values Time [s] Time [s]
but the difference is quite small. The steering-wheel
torque prediction appears to be very accurate, the Figure 8: Chassis and cabin signals during a J-turn
peak values are predicted accurately for negative
values and diverge a little bit for positive values. An
Data 1
important thing to notice here is the sudden jump 10 Prediction
when the steering-wheel motion changes direction.

Tsw [−]
δsw [−]
Bot the measurement and the simulation show a sud- 0
den jump in steering-wheel torque, this is something 0
that the driver definitely notices and is causes by
-1
the friction on the input side of the steering-house. -10
Figure 10 shows the accelerometers and gyroscopes 0 10 0 10
during this experiment. At the start of the experiment
there is some difference between the predicted and
0.2 0.2
the measured signals. This has mainly to do with

δkp,R [−]
δkp,L [−]
the initial conditions of the model and the measure- 0 0
ment. Since the vehicle state before the start of the
experiment is unknown the simulation is started by -0.2 -0.2
a 5 second initialization period where the vehicle
drives with the speed and steering-wheel angle as the 0 10 0 10
experiment starts with. This initialization part is cut
off again after the simulation. Apart from that, the 0.5
lateral accelerations and yaw-rates measured on the 0.2
chassis are predicted accurately. The same holds for

Fdl [−]
δpa [−] 0 0
the lateral acceleration and roll-rate measured in the
cabin. -0.2 -0.5

Figure 11 shows the steering-wheel torque as a -0.4

function of steering-wheel angle during the sinu- 0 10 0 10
soidal input. From this figure it becomes clear that Time [s] Time [s]
there is friction present in the system since change Figure 9: Steering-system signals during a sinusoidal input
of steering-wheel angle direction results in a jump
in torque. Also the non-linearity of the hydraulic
system can be seen in the curved shape of this plot.
The top-right graph shows the lateral acceleration as 1 1
a function of steering-wheel angle. From this graph it Data
can be concluded that the phase of the signal is also Prediction
0.5 0.5
ay,F [−]

ay,R [−]

correctly predicted. The two middle graphs show the

lateral acceleration at the front and the rear of the 0 0
chassis as a function of steering-wheel torque. Again
the non-linearity of the hydraulic system is observed -0.5 -0.5
since the lines are curved. The friction is visible in 0 10 0 10
these graphs as well since the signal jumps horizon-
tally upon changing direction in lateral acceleration. 1 1
The bottom two figures show the yaw-rate front and
rz,F [−]

rz,R [−]

rear as a function of steering-wheel torque. Again

0 0
here the curved shape is seen and the horizontal
jumps show the friction present in the system.
-1 -1
Overall the conclusion in drawn that this steering- 0 10 0 10
system model in combination with the multi-body
truck model and the hub geometry is able to predict all 0.4
measured signals in a satisfactory manner. The angles, 0.5
accelerations, yaw-rates and roll-rates can be pre- 0.2
ay,C [−]

rx,C [−]

dicted with a very high accuracy while the steering- 0 0

wheel torque and the drag-link force appear to be -0.2
harder to predict. -0.5
-0.4

0 10 0 10
Time [s] Time [s]

Figure 10: Chassis and cabin signals during a sinusoidal input

 Data 7 ACKNOWLEDGMENTS
2 0.5
Prediction
This research was supported by the Stichting voor

ay,F [−]
Tsw [−]

1
0
de Technische Wetenschappen (Dutch Technology
0 Foundation) STW.
-1 -0.5
REFERENCES
-10 0 10 -10 0 10
δsw [−] δsw [−] [1] S. Bennett. Heavy duty truck systems. Cengage
0.5 0.5 Learning, 2015.
ay,F [−]

ay,R [−]

[2] I. Besselink, J. Pauwelussen, and B. Kraaijen-

0 0
hagen. Greening and safety assurances of future
modular road vehicles. In VDI Commercial ve-
-0.5 -0.5 hicles 13th International Conference, 2015.
-1 0 1 -1 0 1 [3] A. DellAmico. Pressure control in hydraulic
Tsw [−] Tsw [−] power steering systems. 2013.
1 1
[4] W.-J. Evers, I. Besselink, A. van der Knaap, and
rz,F [−]

rz,R [−]

H. Nijmeijer. Development and validation of a

0 0
-1
-1 0 1
Tsw [−]
Figure 11: Steering-wheel torque as a function of steering-wheel
angle and accelerations and yaw-rates as a function of steering-
wheel angle or torque during a sinusoidal input
5 CONCLUSIONS AND RECOMMENDATIONS
A truck steering-system model has been coupled to a
multi-body tractor semi-trailer model. A steady-state
cornering test with an instrumented vehicle has been
done to estimate the self-aligning stiffness and the
cornering stiffness of the tyres. This instrumented ve-
hicle is also used to perform a J-turn and a 0.4 Hz
sinusoidal input to verify the steering-system model
predictions. The steering-system model in combina-
tion with the multi-body truck model appears to be
able to predict the motion of the vehicle (lateral ac-
celerations, yaw-rates and roll-rate) accurately. The
prediction of the steering-wheel torque and drag-link
force appears to be harder but is also predicted in a
satisfactory manner.
6 FUTURE WORK
Future work will consist out of verification with tests
of a different input frequency. When the model is fully
verified it will be used to investigate the possibilities
of making the steering-system active by means of an
extra actuator. This active steering-system can be used
to analyze different support systems such as a merg-
ing support and a lane keeping/changing assistant.
modular simulation model for commercial ve-
hicles. International Journal of Heavy Vehicle
-1
Systems, 16, 2009.
-1 0 1
[5] J. Loof, I. Besselink, and H. Nijmeijer. Compo-
Tsw [−]
nent based modeling and validation of a steering
system for a commercial vehicle. In 24th IAVSD
- Dynamics of Vehicles on Roads and Tracks,
2015.
[6] J. Loof, I. Besselink, W. Post, and H. Nijmei-
jer. Development of a truck steering system
model including hydraulics to predict the steer-
ing wheel torque. In 9th Graz Symposium Vir-
tual Vehicle, 2016.
[7] P. Mock. European vehicle market statistics,
2013.
[8] P. E. Pfeffer, M. Harrer, and D. Johnston. Inter-
action of vehicle and steering system regarding
on-centre handling. Vehicle System Dynamics,
46(5):413–428, 2008.
[9] M. Rösth. Hydraulic power steering system de-
sign in road vehicles: Analysis, testing and en-
hanced functionality. 2007.
[10] C. Schoon, M. Doumen, and D. de Bruin. The
circumstances of blind spot crashes and short-
and long-term measures, 2008.
[11] H. Summala and T. Mikkola. Fatal accidents
among car and truck drivers: effects of fatigue,
age, and alcohol consumption. Human Fac-
tors: The Journal of the Human Factors and Er-
gonomics Society, 36(2):315–326, 1994.
[12] SWOV. Swov fact sheet. blind spot crashes,
2015.

View publication stats