Mechatronics 78 (2021) 102629

Contents lists available at ScienceDirect

Mechatronics
journal homepage: www.elsevier.com/locate/mechatronics

Steering feel improvement by mathematical modeling of the Electric Power

Steering system✩ , ✩✩
Jung Hyun Choi a , Kanghyun Nam b ,∗, Sehoon Oh a ,∗
a
Department of Robotics Engineering, DGIST, Daegu, 42988, Republic of Korea
b
School of Mechanical Engineering, Yeungnam University, Gyeongsan 38541, Republic of Korea

ARTICLE INFO ABSTRACT

Keywords: Currently, the Electric Power Steering (EPS) system is an essential component of the vehicle because it provides
Electric steering system assistive steering torque to the driver. To ensure a faster steering response, the position of the EPS in some
Steering feel vehicles is moved closer to the tire rather than the steering wheel. The steering torque, which is provided by
Double Universal Joint
the EPS in the steering system, mainly affects the driver’s feel while steering. Therefore, the driver often feels
Torque compensation
uncomfortable owing to such positioning of the EPS in the steering system. In particular, the nonlinearity of
the Universal Joint (UJ), which is one of parts of the steering system, can be felt at the steering wheel side.
In this paper, we proposed an algorithm based on the mathematical model of the steering torque in the
steering system to improve the steering feel. The mathematical model is structured using parameters that can
be obtained from the information of the steering system. Moreover, the formulation of the steering torque
consists of the two parts, namely the deformation part, which describes the propagation inside the steering
system, and the friction part that describes the inherent friction in the UJ.
Simulation and experiments were conducted to verify the proposed mathematical model with similar
conditions to the real tire load during the steering motion. Furthermore, to improve the driver’s feel during
steering, a torque compensation algorithm is proposed and verified through experiments.

1. Introduction achieve the desired torque using a hydraulic device attached to either
the pinion or the rack [1,2]. However, the hydraulic assistive system
The vehicle steering system performs two functions: first, it transfers has several disadvantages: it requires a power device for managing
the road and vehicle state information to the driver through the steering hydraulic pressure, the maintenance of oil pressure is demanding, and
wheel, and second, it changes the driving direction. In addition to these the weight of the hydraulic device is not sufficiently light.
functions, in recent years, the steering feel to the driver has become These drawbacks affect the efficiency of the vehicles, and thus a
an important factor in determining the vehicle quality and the ride different type of steering assistance system called EPS (Electric Power
comfort; therefore, technologies that can improve the steering feel are Steering) has been developed. Nowadays, most vehicles are equipped
required. The structure of the steering system can be divided into three with this device [3,4].
parts: (1) the steering wheel, which interacts with the driver, (2) the From the structural viewpoint, the EPS is largely categorized into
rotating part, which includes the column, the Universal Joint (UJ), and two types [5]: Column type electric power steering (C-EPS) and Rack/
the pinion, and (3) the rack, which includes the tie rod that is connected Pinion type electric power steering (R-EPS/P-EPS). In C-EPS, the steer-
to the wheel. The force on the rack, which is related to the vertical
ing motor is attached to the column close to the steering wheel in a
vehicle load on the tires and the tire friction, is transmitted as the
parallel way to its axis, which is widely applied in vehicles. However,
amount of torque on the steering wheel.
the additional steering torque provided by the assistive motor may
As this required steering torque is very large, a device that as-
be too strong for the latter parts, and thus these parts have to be
sists the steering torque applied by the driver was developed, which
strengthened to prevent them from being twisted.
recognizes the driver’s steering direction and compliments it so as to

✩ This work was supported in parted by the National Research Foundation of Korea (NRF) under Grant. (NRF2019R1A2C2011444) and supported in parted
by Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education (2020R1A6A3A13074159).
✩✩ This paper was recommended for publication by Associate Editor Cong Wang.
∗ Corresponding authors.
E-mail addresses: [email protected] (K. Nam), [email protected] (S. Oh).

https://doi.org/10.1016/j.mechatronics.2021.102629
Received 4 March 2021; Received in revised form 8 May 2021; Accepted 10 July 2021
Available online 29 July 2021
0957-4158/© 2021 Elsevier Ltd. All rights reserved.
J.H. Choi et al. Mechatronics 78 (2021) 102629

In the R-EPS/P-EPS type, the steering motor is installed in parallel

on either the rack side or the pinion axis. Therefore, the load on
the steering devices is relatively small compared to the C-EPS type,
which leads to reduction in the size of the column and the UJs. In
addition, because the motor is installed closer to the wheel, it provides
a faster response time of the steering than that provided by the C-
EPS type. R-EPS/P-EPS is widely applied to luxury sedans owing to
its advantageous steering response. In spite of these merits, R-EPS/P-
EPS has the following limitations; drivers feel oscillatory behaviors
even when the steering wheel is rotating at constant speeds. Moreover,
unequal steering torques are detected for right and left turns around
on-center steering wheel position. This phenomenon is caused by the
inherent characteristic of the UJ and the position of Steering Torque
Sensor (STS) in the EPS. Firstly the transformation of the rotational
motion through the UJ exhibits nonlinear characteristic; in other words,
the relationship between the input motion and the output motion of the
UJ is nonlinear [6]. The STS of an EPS is usually installed between the
steering wheel, and therefore, the relation between the STS and the
steering motor differs according to the type of EPS. There exists a UJ Fig. 1. An illustration of the entire steering system, which includes a double Universal
Joint (UJ), the Steering Angle Sensor (SAS) and the Steering Torque Sensor (STS).
between the STS and the steering motor in the R-EPS/P-EPS, whereas
there is no STS between the UJ and the motor in the C-EPS.
In other words, the nonlinearity of the UJ affects the steering torque
control in R-EPS/P-EPS. Moreover, the friction in the Revolute Joints • Development of the torque compensation algorithm to achieve
(RJs) of the UJ is another factor which increases the nonlinearity. To improved steering feel based on the proposed mathematical model
overcome this undesired phenomenon, two UJs are connected in series,
This paper is organized as follows. In Section 2 mathematical model
e.g., in the traction drive line used in trucks [7]. Nevertheless, these
of the steering torque is derived. Further, to verify it the simulation
nonlinearities deteriorate the steering feel of the vehicle, particularly
performed and the experimental results thus obtained are presented
in the R-EPS/P-EPS.
in Section 3. Section 4 introduces the torque compensation algorithm
The steering torque–steering angle map expression [8] has been
and the corresponding experimental results using the proposed math-
developed and utilized as the guideline of the steering feel for EPS to
ematical model. Finally, Section 5 presents the conclusions of this
realize, and many studies have focused on the precise realization of
study.
the desired steering feel described using this map, by means of EPS
motor control and sensing technologies [9,10]. Other studies have been
conducted focusing on the steering feel of EPS: a quantitative analysis 2. Mathematical model of the Double Universal Joint
was conducted based on measurements of the steering feel of vehicles
to study the comfortable steering feel [11]. In this section, we present a mathematical model of the steering
On the other hand, the nonlinearity issues of the double UJ have system related to the steering feel. To this end, the entire steering sys-
been studied in [12,13], but none of those studies have investigated the tem is analyzed to obtain the motion and torque relationship between
effect of the friction on the steering feel in a theoretical way: several the tire, steering wheel, and the UJs. In particular, the steering torque,
studies have focused on the kinematic relationship while ignoring the which plays a dominant role in the steering feel, was investigated
friction effect inside the UJ. Even though some studies have included analytically.
the friction analysis of the UJ, the results they provided are limited.
For example, the mathematical analysis is missing in some studies [14], 2.1. Framework of the mathematical model of the steering wheel system
whereas in others, the results are too complicated to be applied for the
actual steering system [15,16]. [17] conducted an analysis of the UJ Fig. 1 depicts the complete structure of the Pinion type steering
friction and suggested a control strategy to address it, but the target system, which is the focus of this study. The steering wheel is connected
EPS was only a C-EPS where the combined effect of the friction and to the SAS (Steering Angle Sensor), which measures the steering wheel
the UJs cannot be found. angle. The tire direction is steered through the tie rod which is driven
Therefore, analysis on how the frictions of the UJs appear in the by the rack and pinion, and the EPS is attached in parallel to the pinion
steering torque in addition to the desired steering torque–angle rela- axis to provide an assistive steering torque.
tionship should be conducted precisely, which can be applied for the In particular, the STS is attached in series with the pinion axis
analysis of various types of EPS and the compensation of the frictions. to measure the steering torque. The assistive torque supplied by the
In this study, based on the geometrical analysis of mechanical parts and EPS is controlled utilizing the measurement of the STS to realize the
sensors, we developed the mathematical model of the steering torque predefined impedance model that determines the steering feel.
while taking into consideration the nonlinear characteristic of the UJ. Finally, a double-UJ connects the steering wheel and the pinion axis,
Furthermore, based on this mathematical model, a compensation algo- thus addressing the misalignment issue. Moreover, it also transfers the
rithm for the steering torque to improve the steering feel is proposed. steering wheel motion to the pinion as well as the contacting force of
The contribution points of this paper are summarized as follows: the tire to the steering wheel.
UJ is employed to efficiently transfer motion and torque between
• Mathematical formulation of the double UJ kinematics/ statics the two misaligned axes. Generally, the output motion of the UJ
including joint frictions through the geometric approach suffers from velocity variation due to the nonlinear characteristic of its
• Derivation of the mathematical model of steering torque as the transformation which applies even when the input motion is constant.
function of pinion torque and the UJ friction considering UJ In some applications, two UJs are connected in series to overcome this
formulation undesired distortion, e.g., in the traction drive line used in trucks [7].
• Verification of the proposed mathematical model through precise However, the cancellation of the this distortion requires some special
simulations and experiments using actual UJs conditions to be met: the bend angles of two UJs should be the equal

2
J.H. Choi et al. Mechatronics 78 (2021) 102629

Fig. 2. Geometrical analysis of the single UJ. 𝜃𝐵 and 𝜃𝐹 are the rotation angles of the
base shaft and the follower shaft, respectively. 𝛽 is the bend angle or the angle of the Fig. 3. Configuration of the doulble UJ (a) special configuration (𝛽𝐻 = 𝛽𝑃 and 𝜃𝑜𝑠 )
axles with respect to each other. where there is no velocity variation at the output shaft (b) general configuration which
is discussed in this research.

(𝛽𝐻 = 𝛽𝑃 ), and there should be no offset angle in the middle shaft

(𝜃𝑜𝑠 = 0), as shown in Fig. 3(a). When these conditions are met, the The equation that relates 𝜃𝐵 and 𝜃𝐹 can be expressed in terms of
condition 𝜔𝐻 = 𝜔𝑃 holds true. vectors 𝑋̂ 𝐵 and 𝑋̂ 𝐹 in the 𝑍𝐵 coordinates as follows:
It is not always this well-conditioned double UJ that is included in [ ]
𝑋̂ 𝐵 = cos 𝜃𝐵 , sin 𝜃𝐵 , 0 , (2)
the steering system. More general structure of the double UJ that is [ ]
adopted in the steering system is illustrated in Fig. 3(b), where 𝛽𝐻 ≠ 𝛽𝑃 , 𝑋̂ 𝐹 = − cos 𝛽 sin 𝜃𝐹 , cos 𝜃𝐹 , sin 𝛽 sin 𝜃𝐹 . (3)
and the offset angle is non-zero (𝜃𝑜𝑠 ≠ 0). It is important to note that 𝑋𝐵 and 𝑋𝐹 are always perpendicular
In this general configuration, velocity variation exists in the output because of the constraint 𝑋̂ 𝐵 ⋅ 𝑋̂ 𝐹 = 0, which is imposed by the Spider.
axis; this deteriorates the steering feel. Specifically, the torque on the Therefore, 𝜃𝐵 can be expressed in terms of 𝜃𝐹 and 𝛽 as follows.
pinion measured by the STS is distorted at the steering wheel, which is
tan 𝜃𝐵 (𝑡)
detected by the driver. Moreover, the friction inherent in the Revolute tan 𝜃𝐹 (𝑡) = ,
cos 𝛽
Joints (RJs) of the UJ (as indicated at the bottom of Fig. 1) also affects tan 𝜃𝐵 (𝑡)
the steering feel. 𝜃𝐹 (𝑡) = tan−1 . (4)
cos 𝛽
In this research, we mathematically investigated the torque distribu-
Hence, the angular velocity of the follower in the single UJ can be
tion of the general double UJ. Specifically, we derived the relationship
𝐻 , both of obtained by differentiating (4) as follows.
between the pinion torque 𝑇𝑃 and the steering torque 𝑇𝑎𝑙𝑙
which are illustrated in Fig. 1. cos 𝛽
𝜔𝐹 (𝑡) = 𝜔𝐵 (𝑡) = 𝐴𝜃𝐵 𝜔𝐵 (𝑡) (5)
We analyzed the steering wheel torque 𝑇𝑎𝑙𝑙 𝐻 as the summation of 1 − sin2 𝛽 cos2 𝜃𝐵 (𝑡)
two terms in the RJs, namely, the transmitted pinion torque and the The double UJ is composed of two UJs in series as shown in Fig. 3(a);
friction torque. These two factors should be properly converted in order therefore it consists of three yokes (Yoke𝐻 , Yoke𝑚 and Yoke𝑃 ), and
𝐻 in the steering wheel axis coordinates as follows.
to express 𝑇𝑎𝑙𝑙 the relationship among their three angular velocities 𝜔𝐻 , 𝜔𝑚 , and 𝜔𝑃
𝐻 should be analyzed. Here, 𝛽𝐻 and 𝛽𝑃 are the bending angles between
𝑇𝑎𝑙𝑙 (𝑡) = 𝑇𝑃𝐻 (𝑡) + 𝑓 𝐻 (𝑡)
Yoke𝐻 and Yoke𝑚 and between Yoke𝑚 and Yoke𝑃 , respectively.
= 𝐴𝑘 𝑇𝑃 (𝑡) + 𝑓 𝐻 (𝑡), (1) The middle shaft has two yokes, Yoke𝐻,𝑚 and Yoke𝑚,𝑃 , which can
be offset as shown in Fig. 3(b), where 𝜃𝑜𝑠 is the twist angle between
where 𝐴𝑘 is the static relationship between the pinion torque and
Yoke𝐻,𝑚 and Yoke𝑚,𝑃 . Note that Fig. 3(a) is a special configuration of
the steering wheel torque, which is determined by the transformation
the double UJ, wherein 𝜃𝑜𝑠 is set to zero, and thus two axes of Yoke𝐻
characteristic of the double UJ. 𝑓 𝐻 (𝑡) indicates the effect of the friction
and Yoke𝑃 intersect at a certain point. In steering systems, various
torque due to the RJs.
double UJ configurations are utilized, and therefore the general con-
figuration of double UJ, as depicted in Fig. 3(b) should be considered
2.2. Kinematic and static analysis of double UJ through geometrical ap- with a non-zero 𝜃𝑜𝑠 .
proach Based on these considerations, the relationship between 𝜔𝐻 , 𝜔𝑚 ,
and 𝜔𝑃 can be expressed as follows.
To obtain 𝐴𝑘 in (1), the kinematics of the UJ was analyzed based cos 𝛽𝐻
on the geometrical features. In a UJ, two yokes (Yoke𝐵 and Yoke𝐹 𝜔𝑚 (𝑡) = 𝜔𝐻 (𝑡) = 𝐴𝐻,𝑚 (𝑡)𝜔𝐻 (𝑡), (6)
1 − sin2 𝛽𝐻 cos2 𝜃𝐻 (𝑡)
in Fig. 2) are connected via the Spider as illustrated in Fig. 2. Three
cos 𝛽𝑃
dimensional relationship between two yokes can be described using the 𝜔𝑃 (𝑡) = 𝜔𝑚 (𝑡) = 𝐴𝑚,𝑃 (𝑡)𝜔𝑚 (𝑡). (7)
two planes of rotation 𝑍𝐵 and 𝑍𝐹 , which are perpendicular to the axis 1 − sin2 𝛽𝑃 cos2 (𝜃𝑚 (𝑡) + 𝜃𝑜𝑠 )
of Yoke𝐵 and Yoke𝐹 , respectively. Finally, the resultant angular velocity relationship of the double UJ is
The angle between 𝑍𝐵 and 𝑍𝐹 , which is denoted as 𝛽, determines (8) obtained by combining (6) and (7) and is given as follows:
the nonlinear transformation characteristic between 𝜃𝐵 and 𝜃𝐹 . 𝑋{𝐵,𝐹 }
𝜔𝑃 (𝑡) = 𝐴𝑚,𝑃 (𝑡)𝐴𝐻,𝑚 (𝑡)𝜔𝐻 (𝑡). (8)
and 𝑌{𝐵,𝐹 } denote the coordinate axes on the planes 𝑍{𝐵,𝐹 } , respec-
tively. As Yoke𝐵 rotates at the angle 𝜃𝐵 , the vector 𝑋𝐵 rotates on the The double UJ transmits not only the motions 𝜔𝑃 and 𝜔𝐻 but also
plane 𝑍𝐵 in the direction given by the unit vector 𝑋̂ 𝐵 . Simultaneously the corresponding torques 𝑇𝑃 and 𝑇𝐻 , as illustrated in Fig. 4. The statics
Yoke𝐹 rotates on the plane 𝑍𝐹 at an angle 𝜃𝐹 to 𝑋̂ 𝐵 , as shown in Fig. 2. between the pinion torque 𝑇𝑃 and the steering torque 𝑇𝐻 can be derived

3
J.H. Choi et al. Mechatronics 78 (2021) 102629

Fig. 6. The complete process required to calculate the steering wheel torque by using
the sensor information and the configuration of the steering system.

Fig. 5 illustrates the angles of all the RJs namely, 𝜃𝑅𝐽 1 , 𝜃𝑅𝐽 2 , 𝜃𝑅𝐽 3 ,
and 𝜃𝑅𝐽 4 , which are the rotation angles along 𝑋{𝐵,𝐹 } and 𝑌{𝐵,𝐹 } in
Fig. 2.
Fig. 4. Motion and torque transfer through the double UJ in the steering system. Firstly, the RJ angles 𝜃𝑅𝐽 {1,2} of the first UJ are determined as

𝜃𝑅𝐽 1 (𝑡) = 𝛽𝐻 sin 𝜃𝐻 (𝑡), (11)

𝜃𝑅𝐽 2 (𝑡) = 𝛽𝐻 cos 𝜃𝐻 (𝑡), (12)

where, 𝜃𝐻 is the steering wheel angle. This relationship leads to the

restriction −𝛽𝐻 < 𝜃𝑅𝐽 {1,2} < 𝛽𝐻 , which is implied on the RJ angles.
Following this relationship, the angular velocities of RJ1 and RJ2
are determined as follows.

𝜔𝑅𝐽 1 (𝑡) = 𝛽𝐻 cos 𝜃𝐻 (𝑡)𝜔𝐻 (𝑡), (13)

𝜔𝑅𝐽 2 (𝑡) = −𝛽𝐻 sin 𝜃𝐻 (𝑡)𝜔𝐻 (𝑡). (14)

In this paper, the friction of 𝑖th RJ is denoted as 𝑓𝑅𝐽 𝑖 (𝑡), and

Coulomb friction is assumed to be the most dominant friction at each
joint. This is ascertained by the fact that the steering torque is measured
Fig. 5. Revolution Joints (RJs) in the UJ and the definition of the angle of each RJ. when the wheel is rotated at constant speeds, and the level of the
measured steering torques are almost same regardless of speeds (30,
50, 150 rpm). Therefore the friction force is expressed as 𝑓𝑅𝐽 𝑖 (𝑡) =
based on this kinematic relationship and the virtual work principle, −|𝑅𝐽 𝑖|sign(𝜔𝑅𝐽 𝑖 (𝑡)).
𝐻 𝜔 = 0, which is given as
i.e., 𝑇𝑃 𝜔𝑃 − 𝑇𝑎𝑙𝑙 The kinematic constraint of the RJs specifies the way how the
𝐻
frictions are reflected on the steering wheel. Hence, the virtual work
𝐻
𝑇𝑎𝑙𝑙 (𝑡)𝜔𝐻 (𝑡) = 𝑇𝑃 (𝑡)𝜔𝑃 (𝑡) 𝐻
principle, 𝑓𝑅𝐽 {1,2} 𝜔𝑅𝐽 {1,2} − 𝑓𝑅𝐽 𝜔 = 0 can be applied to derive the
{1,2} 𝐻
𝜔 (𝑡) 𝐻 friction force transmission relationship as follows.
𝑇𝑃 (𝑡) = 𝐻 𝑇𝑎𝑙𝑙 (𝑡) = 𝐴−1 −1 𝐻
𝑚,𝑃 (𝑡)𝐴𝐻,𝑚 (𝑡)𝑇𝑎𝑙𝑙 (𝑡). (9)
𝜔𝑃 (𝑡) 𝐻
𝑓𝑅𝐽 {1,2}
(𝑡)𝜔𝐻 (𝑡) = 𝑓𝑅𝐽 {1,2} (𝑡)𝜔𝑅𝐽 {1,2} (𝑡),
The above equation describes how the steering wheel torque 𝑇𝑎𝑙𝑙 𝐻 is
𝐻
𝜔𝑅𝐽 {1,2} (𝑡)
transferred to the pinion, as 𝑇𝑃 (𝑡). 𝑓𝑅𝐽 {1,2}
(𝑡) = 𝑓𝑅𝐽 {1,2} (𝑡). (15)
𝜔𝐻 (𝑡)
Similarly the means through which the pinion torque 𝑇𝑃 , which
is generated by the interaction between the tire and the road, is Substituting (13) and (14) into (15) gives us the steering torque
transferred to the steering wheel can be derived. It should be noted that caused by the friction of RJs given as follows.
this transformation is very significant when considering the steering 𝐻
𝑓𝑅𝐽 1
(𝑡) = 𝛽𝐻 cos 𝜃𝐻 (𝑡)𝑓𝑅𝐽 1 (𝑡), (16)
feel. Eq. (10) describes the steering torque 𝑇𝑃𝐻 , which is transmitted
𝐻
from 𝑇𝑃 . 𝑓𝑅𝐽 2
(𝑡) = −𝛽𝐻 sin 𝜃𝐻 (𝑡)𝑓𝑅𝐽 2 (𝑡). (17)

𝑇𝑃𝐻 𝜔𝐻 (𝑡) = 𝑇𝑃 (𝑡)𝜔𝑃 (𝑡) Similar to (11) and (12), kinematics of the angular velocity of RJ3 and
RJ4 can be derived as
𝑇𝑃𝐻 (𝑡) = 𝐴−1 −1
𝑚,𝐻 (𝑡)𝐴𝑃 ,𝑚 (𝑡)𝑇𝑃 (𝑡) = 𝐴𝑘 𝑇𝑃 (𝑡), (10)
𝜃𝑅𝐽 3 (𝑡)=𝛽𝑃 sin(𝜃𝑚 (𝑡)+𝜃𝑜𝑠 ) = 𝛽𝑃 sin 𝜃𝑚+𝑜𝑠 (𝑡) (18)
where
𝜃𝑅𝐽 4 (𝑡)=𝛽𝑃 cos(𝜃𝑚 (𝑡)+𝜃𝑜𝑠 ) = 𝛽𝑃 cos 𝜃𝑚+𝑜𝑠 (𝑡) (19)
1 − sin2 𝛽𝐻 cos2 (𝜃𝑚 (𝑡) − 𝜃𝑜𝑠 )
𝐴−1
𝑚,𝐻 (𝑡) = ,
cos 𝛽𝐻 where, 𝜃𝑚+𝑜𝑠 is the rotation angle of Yoke𝐻,𝑚 in Fig. 3(b), which consists
1 − sin2 𝛽𝑃 cos2 𝜃𝑃 (𝑡) of the rotation angle 𝜃𝑚 and the offset angle between Yoke𝐻,𝑚 and
𝐴−1
𝑃 ,𝑚 (𝑡) = . Yoke𝑚,𝑃 .
cos 𝛽𝑃
By differentiating (18) and (19), the angular velocities of RJ3 and
2.3. Friction analysis of the double universal joint RJ4 can be represented in terms of 𝜔𝑚 (𝑡) as follows.

𝜔𝑅𝐽 3 (𝑡) = 𝛽𝑃 cos 𝜃𝑚+𝑜𝑠 (𝑡)𝜔𝑚 (𝑡), (20)

The steering feel is also effected by the inherent friction in the RJs
(RJ1, RJ2, RJ3, RJ4) of the UJ. The steering feel is further complicated 𝜔𝑅𝐽 4 (𝑡) = −𝛽𝑃 sin 𝜃𝑚+𝑜𝑠 (𝑡)𝜔𝑚 (𝑡). (21)
by the friction in each RJ, which is transmitted to the steering feel
𝜔𝑅𝐽 {3,4} (𝑡) can also be related to the steering wheel velocity 𝜔𝐻 (𝑡) by
through the UJ structure. In other words, the effects of these frictions
substituting (6) in (20) and (21).
become oscillatory even if the frictions themselves are mainly of the
type Coulomb friction, which is generally static in nature. Furthermore, 𝜔𝑅𝐽 3 (𝑡) = 𝛽𝑃 cos 𝜃𝑚+𝑜𝑠 (𝑡)𝐴𝐻,𝑚 (𝑡)𝜔𝐻 (𝑡), (22)
to define this relationship, we performed a kinematic analysis to obtain
𝜔𝑅𝐽 4 (𝑡) = −𝛽𝑃 sin 𝜃𝑚+𝑜𝑠 (𝑡)𝐴𝐻,𝑚 (𝑡)𝜔𝐻 (𝑡). (23)
the relationship of the angular velocities of RJs with respect to the
steering wheel motion. Subsequently, the relationship between the As for 𝑓𝑅𝐽 {1,2} , it can be assumed the Coulomb friction mainly
friction of RJs and the steering torque is derived. affects 𝑅𝐽 {3, 4} which is represented as 𝑓𝑅𝐽 {3,4} (𝑡) = −|𝑅𝐽 {3, 4}|

4
J.H. Choi et al. Mechatronics 78 (2021) 102629

Fig. 7. Verification strategy of the mathematical model.

𝑠𝑖𝑔𝑛(𝜔𝑅𝐽 {3,4} (𝑡)). The amount of the frictions 𝑓𝑅𝐽 {3,4} transmitted on the
steering wheel can be calculated utilizing the kinematic relationship
derived above.
𝐻
𝑓𝑅𝐽 3
(𝑡) = 𝐴−1
𝐻,𝑚 (𝑡)𝛽𝑃 cos 𝜃𝑅𝐽 3 (𝑡)𝑓𝑅𝐽 3 (𝑡), (24)
𝐻
𝑓𝑅𝐽 4
(𝑡) = 𝐴−1
𝐻,𝑚 (𝑡)𝛽𝑃 (− sin 𝜃𝑅𝐽 4 (𝑡))𝑓𝑅𝐽 4 (𝑡). (25)
Fig. 8. Verification Environments (a) Experimental setup to verify the proposed math-
Finally, the effect of all the RJ frictions on the steering wheel torque ematical model of the steering system (b) Simulation model based on the Simscape™in
can be summarized as the summation of (16), (17), (24) and (25), as the Matlab©/Simulink© environment.
given below.
∑
4
𝑓 𝐻 (𝑡) = 𝐻
(26) was verified through the experimental results with the pinion torque
𝑓𝑅𝐽 𝑛 (𝑡)
𝑛=1 𝑇𝑃 set to 0. Finally, a total wheel torque was measured with loads on
the pinon to mimic the rack force 𝐹𝑟𝑎𝑐 [18], and the steering torque
2.4. Calculation of steering torque using proposed model thus measured was compared with the model results of (27), which
represent the results of the proposed model.
The complete mathematical model for the steering torque that is Fig. 8(a) illustrates the experimental apparatus, which consists of
affected by the pinion torque given in (10) and the RJ frictions specified following components.
in (26) can be combined as shown in (27).
Based on this modeling, the 𝑇𝑎𝑙𝑙 𝐻 in (27) be estimated in real-time. • Two torque sensors (STS) and encoders (SAS) for measuring the
Fig. 6 illustrates the estimation process, wherein 𝑇𝑃 (𝑡) is measured us- torque/angle of the pinion and the steering wheel axis.
ing STS, and 𝜃𝑚 (𝑡), 𝜃𝑅𝐽 {1,2,3,4} can be calculated utilizing 𝜃𝐻 (𝑡) measured • One variable electric brake at the pinion axis to mimic the rack
by SAS. force.
𝜃𝑜𝑠 can be calculated using 𝜃𝑅𝐽 3 (0) in (18), which is the angle • One motor at the steering wheel axis to generate the rotary
when the steering wheel is at the on-center. Moreover, 𝛽𝐻 and 𝛽𝑃 motion to the mimic driver’s motion.
are the initial design parameters, which can be obtained from the • One motor at the pinion axis to compensate for the steering wheel
manufactures. torque fluctuation due to nonlinearity and friction.
• A dual universal joint with various friction conditions.
∑
4
𝐻
𝑇𝑎𝑙𝑙 (𝑡) = 𝑇𝑃𝐻 (𝑡) + 𝑓 𝐻 (𝑡) = 𝐴𝑘 𝑇𝑃 (𝑡) + 𝐻
𝑓𝑅𝐽 𝑛 (𝑡), Because the target system of this research was the R-EPS, a EPS
𝑛=1
2 2
motor was installed on the pinion side to provide the assistive steering
1 − (sin 𝛽𝐻 cos2
(𝜃𝑚 (𝑡) − 𝜃𝑜𝑠 )) 1 − sin 𝛽𝑃 cos2 𝜃𝑃 (𝑡) torque. Moreover, in this study, this motor was also utilized to com-
= 𝑇𝑃 (𝑡)
cos 𝛽𝐻 cos 𝛽𝑃 pensate for the effect of the nonlinearity of the UJ and its friction. The
+ 𝛽𝐻 cos 𝜃𝑅𝐽 1 (𝑡)𝑓𝑅𝐽 1 (𝑡) + (−𝛽𝐻 sin 𝜃𝑅𝐽 2 (𝑡))𝑓𝑅𝐽 2 (𝑡) torque sensor STS was attached on the pinion axis, and the angle sensor
+ 𝐴−1 SAS was attached on the steering wheel axis. The measurements of STS
𝐻,𝑚 (𝑡)𝛽𝑃 cos 𝜃𝑅𝐽 3 (𝑡)𝑓𝑅𝐽 3 (𝑡) 𝐻 , as illustrated in Fig. 6.
and SAS were utilized to calculate 𝑇𝑎𝑙𝑙
+ 𝐴−1
𝐻,𝑚 (𝑡)𝛽𝑃 (− sin 𝜃𝑅𝐽 4 (𝑡))𝑓𝑅𝐽 4 (𝑡). (27) Moreover, the steering wheel axis torque sensor and the pinion axis
angle sensor were added to verify the mathematical model proposed in
3. Verification of the steering wheel torque mathematical model this paper. These sensors were used only for the verification and were
not utilized in the actual vehicle steering system.
In this section, the verification of the proposed steering wheel Furthermore, an electric motor was installed to generate patterned
torque model in (27) through simulations and experiments is presented. steering wheel motion, and an electric brake was attached on the pinion
The model consists of two parts — kinematic relationship 𝐴𝑘 (𝑡) and to mimic the rack force.
the friction effect 𝑓 𝐻 (𝑡). Accordingly the experimental verification is Because the measurements in the experiments could not measure
conducted in three steps, as shown in Fig. 7. Firstly, to verify the the details of each component, simulations were conducted to verify
kinematic relationship in (8), the angular velocity of the pinion axis these details of friction obtained using the proposed model. Fig. 8(b)
was measured with a certain steering wheel motion pattern, which was shows the Matlab© /Simulink© simulation environment of the steering
then compared with the calculation output obtained using (6), (7) and system. The physical model of the steering system was first drawn using
(8). Secondly, the effect of friction on the steering wheel 𝑓 𝐻 in (26) the 3D CAD tool and was then imported to Simulink© using Simscape™.

5
J.H. Choi et al. Mechatronics 78 (2021) 102629

Fig. 9. Verification of the kinematic model: (a) Experimental condition of the double
UJ (b) Experimental results.

3.1. Verification of kinematic model

The kinematic model, 𝐴𝑘 , was verified by comparing the angular

velocities of the pinion axis calculated from (8) and the actual ex-
perimental measurements. In the experiment, the steering wheel was
rotated left and right rotate at the constant speed of 30 rpm; from the
on-center to −540◦ then to 540◦ and back to the on-center, as shown
in Fig. 9(a).
The results are plotted in Fig. 9(b) where the solid black, solid
cyan, and dashed blue lines indicate the angular velocity of the steering
wheel, the experimental results, and the equation of the pinion axis,
Fig. 10. Verification of RJ motion and torque through the Simcape™Model (a)
respectively. Note that the experimental values of the steering wheel
Verification of the motion model between steering wheel and RJ1,RJ2,RJ3, and RJ4 (b)
angular velocity were used for the 𝜔𝐻 in (8). Verification of the friction model between steering wheel and RJ1,RJ2,RJ3, and RJ4 (c)
As expected, the oscillation of the pinion axis was observed even Condition of the double UJ in the friction experiment (d) Steering torque verification
when the steering wheel was rotated at a constant speed. The results of the friction using (26).
of the equation and the experiment were approximately equal with
0.12 Root Mean Square Error (RMSE), which verifies that the proposed
kinematic model can depict the kinematic relationship satisfactorily. It is interesting that the magnitudes of the friction forces transmitted
Moreover, the simulation model, which represents the actual UJ was to the steering torque are smaller than the actual friction forces in
verified as shown by the magenta solid line in Fig. 9(b), which is similar the RJs. This phenomenon can be also verified using (16), (17), (24),
to the experimental result. and (25) between the simulation and calculation results, the calculated
values of RMSEs of the steering torque were 0.002, 0.002, 0.006, and
3.2. Verification of the friction torque model 0.007, respectively, which verifies the accuracy of the proposed model.
Finally, the steering torque by frictions in all the RJs are depicted
We verified the friction effect of RJs to the steering wheel torque. Fig. 10(c). The simulation result obtained using this friction condition
First, the angular velocities of RJs with respect to the steering wheel is plotted in Fig. 10(d) and is compared with the calculation results
angular velocity in (13), (14), (22) and (23) are verified. The effects of (26). The pattern of the steering torque (solid yellow line) was
of the RJ frictions on the steering wheel torques given by (16), (17), observed to be periodic, and the magnitude of the summed friction is
(24), and (25) were verified. The verification of the angular velocity approximately equal to the calculation result.
relationship was conducted only through simulation, because it was In the experiment, various UJs with different friction of RJ were
difficult to measure the angular velocities of the RJs experimentally. used to verify the results of (26). Specifically, for calculating the steer-
To verify the kinematics the steering wheel was rotated at a constant ing torque using (26), the value of friction of each RJ is required and
speed of 5 rad/s (solid black line in Fig. 10) from the on-center to the UJ manufacturer provides rough Coulomb friction values. Fig. 11
540◦ . All the solid lines indicate the simulation results, and all the graphs the experimental results of the steering torque for four different
dashed lines indicate the results of the modeling. Fig. 10(a) plots UJ samples labeled as UJ#1, UJ#2, UJ#3, and UJ#4. The corresponding
the comparison between the simulation results and the model output RMSE values obtained from experimental results and equations (dashed
obtained using (13), (14), (22), and (23). The oscillatory behavior can green line) are 0.044 N m, 0.087 N m, 0.051 N m, and 0.088 N m,
be observed in the RJs even when the steering wheel is rotated at a respectively. Thusly, friction model of the UJ was verified.
constant speed. The angular velocity of each RJ is observed to be lower
than the actual steering wheel speed (5 rad/s). Moreover it can be
observed that velocities calculation performed using (13), (14), (22), 3.3. Verification of the combined mathematical model
and (23) were approximately equal to the velocities obtained from
simulations. The corresponding RMSE values for these velocities were To verify the combined model expressed in (27), which is complete
calculated as 0.23, 0.21, 0.32, and 0.33, respectively, which validates mathematical model proposed in this paper, the experimental condi-
that the proposed RJ velocity model is sufficiently accurate. tions were set as illustrated in Fig. 12(a). The combined load (constant
Fig. 10(b) graphs the simulation results of the four different steering and elastic) was applied to the electric brake to ensure that it imitated
torques, where the steering wheel is rotated similarly to that done in the the rack and pinion system in the vehicle. Note that the experimental
case of Fig. 10(a). The RJ friction in this simulation was considered to values of the pinion torque were used for 𝑇𝑃 (𝑡) in (27) to compare
be Coulomb friction with an amplitude of 0.1 N s/m (solid black line). the steering torque under the same conditions. Other experimental
conditions, including motion and speed of the steering wheel, are same

6
J.H. Choi et al. Mechatronics 78 (2021) 102629

Fig. 11. Experimental results for various UJ samples with different RJ friction values.
The steering wheel motion was set to 30 rpm and under no load condition at the pinion
axis.
Fig. 13. The torque compensation strategy for improving the steering feel in the
steering system.

Fig. 12. Experimental results for comparison between the steering torque obtained Fig. 14. Torque compensation experimental results (a) Torque compensation for
from Eq. (27) and the experiment.(a) Experimental condition (b) Verification of the linearization of the steering wheel torque (b) Torque compensation to eliminate friction
whole mathematical model of the UJ with constant and elastic load at the pinion axis. inside of UJ.
The RMS error was observed to be 0.23 N m.

mathematical model proposed in this paper. The properties of the 𝑇𝑃∗

as those applied in Section 3.1. Experimental results are plotted in are: (1) linearization of the non-linearity between the steering torque
Fig. 12(b) where solid black, solid cyan, and dashed blue lines represent and pinion torque by using 𝐴−1 , and (2) elimination of the friction
𝑘
measured pinion torque, measured steering torque, and the steering
effect using 𝑓 𝐻 as follows:
torque calculated using (27). The RMSE was determined to be 0.23 N m.
Thusly the proposed mathematical model was verified. 𝑇𝑃∗ (𝑡) = 𝐴−1 ∗
𝑘 (𝑡)(𝑇𝐻 (𝑡) − 𝑓𝐻 (𝑡))

= 𝐴−1 ∗ −1
𝑘 (𝑡)𝑇𝐻 (𝑡) − 𝐴𝑘 (𝑡)𝑓𝐻 (𝑡). (29)
4. Steering feel improvement control based on the mathematical
model
4.2. Torque compensation experimental result
4.1. Steering feel improvement by torque compensation
To verify the torque compensation algorithm derived in (29), an
It can be observed from Fig. 12(b) that the steering torque is
experiment was conducted with the combined load applied to the
oscillatory and unbalanced at the on-center which is uncomfortable for
electric brake to mimic the rack and pinion condition in the vehicle.
the driver. A torque compensation algorithm that is to be run in the
Other experimental conditions are the same as in Section 3.1. Results
EPS to improve this undesired steering feel, is proposed in this paper.
are plotted in Fig. 14(a) where dashed cyan, solid magenta, and solid
Fig. 13 describes the whole torque compensation algorithm in the 𝐻∗ , torque
blue lines represent the assumed desired steering torque 𝑇𝑎𝑙𝑙
steering system. In the conventional EPS algorithm, 𝑇𝐸𝑃 𝑆 is calculated
𝐻∗ which is the desired steering torque, and 𝑇
by using 𝑇𝑎𝑙𝑙 results obtained without compensation, and those obtained with com-
𝑆𝑇 𝑆 , which is
the measured torque at the pinion, as follows: pensation, respectively. From these results, it was observed that the
oscillation phenomenon was reduced significantly due to the torque
𝐻∗
𝑇𝐸𝑃 𝑆 (𝑡) = 𝐶𝐹 𝐵 (𝑇𝑎𝑙𝑙 (𝑡) − 𝑇𝑝 (𝑡)), (28) compensation; moreover, the steering torque was approximately equal
where, 𝐻∗
𝑇𝑎𝑙𝑙 is the desired steering torque which is same as and is 𝑇𝑃∗ , to the desired one, due to which the steering feel was improved. During
predefined through the 𝑇𝐻 -Map basing on the vehicle states, which are torque compensation, the effectiveness of friction elimination was also
described using parameters such as steering angle (𝜃𝐻 ), steering speed verified as plotted in Fig. 14(a). Because the no load condition (𝑇𝑟𝑎𝑐 = 0)
(𝜃̇ 𝐻 ), rack force(𝐹𝑟𝑎𝑐 ), and vehicle speed (𝑣𝑥 ) [19]. was applied to the electric brake in this experiment, the desired steering
𝐻 𝐻∗ was also zero.
torque 𝑇𝑎𝑙𝑙
However, as mentioned in Introduction, the steering torque 𝑇𝑎𝑙𝑙
is different from the pinion torque 𝑇𝑃 because of the UJ that is in- Therefore (29) can be rewritten as 𝑇𝑃∗ (𝑡) = −𝐴−1
𝑘
(𝑡)𝑓𝐻 (𝑡). The result
stalled between the STS and the steering wheel. To compensate for of friction elimination is effective as depicted in the figure using a
this difference, the desired pinion torque 𝑇𝑃∗ is modified based on the comparison between the solid blue line and the solid magenta line.

7
J.H. Choi et al. Mechatronics 78 (2021) 102629

5. Conclusion [13] Hayama Y. Dynamic analysis of forces generated on inner parts of a double
offset constant velocity universal joint (DOJ): non-friction analysis. Tech. rep.,
SAE Technical Paper; 2001.
In the steering system, the double Universal Joint (UJ) causes
[14] Sun G, Ren W, Zhang J. Virtual product development for an automotive universal
oscillatory and undesired torque in the steering wheel. To address this joint. Int J Automot Technol 2011;12(2):299–305.
problem, we proposed a mathematical model of the steering system [15] Sheu P-P, Chieng W-H, Lee A-C. Modeling and analysis of the intermediate shaft
to improve the steering feel by compensating the oscillatory steering between two universal joints. J Vib Acoust 1996;118(1):88–99.
torque and friction. [16] Biancolini M, Brutti C, Pennestri E, Valentini P. Dynamic, mechanical efficiency,
and fatigue analysis of the double cardan homokinetic joint. Int J Veh Des
Based on the geometrical approach of the double UJ, the equation 2003;32(3–4):231–49.
of the double UJ was derived; it consists of two parts: (1) the kinematic [17] Yang Z-X. Oscillation in electric power steering test torque due to universal joint
part, which transfers the torque from the pinion to the steering wheel, angle and control strategy. J Dyn Syst Meas Control 2013;135(5).
and (2) the friction part, which describes friction of the RJs of the UJ. [18] Abe M. Vehicle handling dynamics: Theory and application. Butterworth-
Heinemann; 2015.
Moreover, the proposed mathematical model including the formulation
[19] Zheng H, Zhou J, Li B. Design of adjustable road feeling performance for
of UJ was arranged by considering its application to the embedded steering-by-wire system. SAE Int J Veh Dyn, Stab, NVH 2018;2(2):121–34.
system with available sensors installed in the steering system. Three
different verification strategies were introduced by using a simulation
model and an experimental setup. Results show that the proposed math- Jung Hyun Choi (S’19) received the B.S. and M.S. de-
ematical model satisfactorily describes the steering torque in various grees in mechanical engineering from Yeungnam University,
experimental conditions. Furthermore, the improvement of the steering Korea, in 2010, and 2013, respectively, and the Ph.D.
feel achieved by implementing the proposed mathematical model was degree in the Department of Robotics Engineering of Daegu
Gyeongbuk Institute of Science and Technology (DGIST),
confirmed through the torque compensation experiments.
Daegu, Korea, in 2021.
From 2013 to 2016, he was a Researcher at the IoT
CRediT authorship contribution statement and Robotics Research Division of DGIST. Since 2021, he
has been an postdoctoral researcher at the DGIST, Daegu,
Jung Hyun Choi: Conceptualization, Methodology, Validation, Korea. His research interests include the development of
mobile robot control algorithms, mechanical design, motion
Writing – original draft, Visualization. Kanghyun Nam: Investigation, control.
Formal analysis, Supervision. Sehoon Oh: Writing – review & editing.

Declaration of competing interest

Kanghyun Nam (S’10–M’12) received the B.S. degree in
mechanical engineering from Kyungpook National Univer-
The authors declare that they have no known competing finan- sity, Daegu, Republic of Korea, in 2007, the M.S.degree
cial interests or personal relationships that could have appeared to in mechanical engineering from Korea Advanced Institute
influence the work reported in this paper. of Science and Technology, Daejeon, Republic of Korea, in
2009, and the Ph.D.degree in electrical engineering from
The University of Tokyo, Tokyo, Japan, 2012.
References
From 2012 to 2015, he was a Senior Engineer with Sam-
sung Electronics Co., Ltd., Gyeonggi-do, Republic of Korea.
[1] Kemmetmuller W, Muller S, Kugi A. Mathematical modeling and nonlinear Since 2015, he has been an Assistant Professor with the
controller design for a novel electrohydraulic power-steering system. IEEE/ASME School of Mechanical Engineering, Yeungnam University,
Trans Mechatronics 2007;12(1):85–97. Gyeongbuk, Republic of Korea. His research interests in-
[2] Dell’Amico A, Krus P. Modeling, simulation, and experimental investigation clude motion control, vehicle dynamics and control, electric
of an electrohydraulic closed-center power steering system. IEEE/ASME Trans vehicles.
Mechatronics 2015;20(5):2452–62. Prof. Nam is a member of the Korean Society of Auto-
[3] Choi H-R, Choe G-H. A multiobjective parametric optimization for passenger-car motive Engineers. He received the Best Paper Award from
steering actuator. IEEE Trans Ind Electron 2010;57(3):900–8. the IEEE T RANSACTION ON I NDUSTRIAL E LECTRONICS
[4] Kim W, Son YS, Chung CC. Torque-overlay-based robust steering wheel angle in 2014.
control of electrical power steering for a lane-keeping system of automated
vehicles. IEEE Trans Veh Technol 2015;65(6):4379–92.
[5] Hur J. Characteristic analysis of interior permanent-magnet synchronous mo-
tor in electrohydraulic power steering systems. IEEE Trans Ind Electron
Sehoon Oh (S’05–M’06–SM’16) received the B.S., M.S., and
2008;55(6):2316–23.
Ph.D. degrees in electrical engineering from The University
[6] Seherr-Thoss H-C, Schmelz F, Aucktor E. Universal joints and driveshafts:
of Tokyo, Tokyo, Japan, in 1998, 2000, and 2005, respec-
analysis, design, applications. Springer Science & Business Media; 2006.
tively.
[7] Mazzei Jr AJ. Passage through resonance in a universal joint driveline system.
He was a research associate at The University of Tokyo
J Vib Control 2011;17(5):667–77.
until 2012, a Visiting Researcher at the University of Texas
[8] Salaani MK, Heydinger GJ, Grygier PA. Experimental steering feel performance
at Austin from 2010 to 2011, a Senior Researcher at the
measures. SAE Trans 2004;665–79.
Samsung Heavy Industries, and a Research Professor at
[9] Balachandran A, Gerdes JC. Designing steering feel for steer-by-wire vehicles
Sogang University. He is currently an Assistant Professor
using objective measures. IEEE/ASME Trans Mechatronics 2014;20(1):373–83.
at the DGIST, Daegu, South Korea. His research interests
[10] Lee MH, Ha SK, Choi JY, Yoon KS. Improvement of the steering feel of an
include the development of force control for human-friendly
electric power steering system by torque map modification. J Mech Sci Technol
motion control algorithms and assistive/exercise devices for
2005;19(3):792–801.
people.
[11] Pfeffer PE, Harrer M, Johnston D. Interaction of vehicle and steering system
Dr. Oh received the Best Transactions Paper Award from
regarding on-centre handling. Veh Syst Dyn 2008;46(5):413–28.
the IEEE Transactions on Industrial Electronics in 2013.
[12] Petrescu FI, Petrescu RV. The structure, geometry, and kinematics of a universal
joint. Indep J Manag Prod (Ijm&P) 2019;10(8).