You are on page 1of 18

Mechanism and Machine Theory 114 (2017) 20–37

Contents lists available at ScienceDirect

Mechanism and Machine Theory


journal homepage: www.elsevier.com/locate/mechmachtheory

Research paper

Improvement of step-climbing capability of a new mobile


robot RHyMo via kineto-static analysis
Dongkyu Choi a,1, Youngsoo Kim a,1, Seungmin Jung a, Hwa Soo Kim b,∗,
Jongwon Kim a,∗
a
School of Mechanical and Aerospace Engineering, Seoul National University, Republic of Korea
b
Department of Mechanical System Engineering, Kyonggi University, Republic of Korea

a r t i c l e i n f o a b s t r a c t

Article history: This study presents the improvement of step-climbing capability of a new mobile robot
Received 17 January 2016 called as RHyMo via kineto-static analysis. To this end, the kinematic analysis of the
Revised 22 March 2017
RHyMo is first performed to obtain the trajectory of its center of mass (CM) when it climbs
Accepted 31 March 2017
up a step. Based on this CM trajectory, the linear/angular speeds and accelerations of its
links are sequentially calculated. Different from the quasi-static analysis, the linear/angular
Keywords: accelerations are combined with the dynamic equations describing the force and moment
Kineto-static analysis equilibriums of RHyMo climbing the step, which are optimally solved by introducing the
Required friction coefficient concept of the required friction coefficient in this study. This kineto-static analysis is re-
Torque analysis peatedly performed by changing the contact angle between the front track of RHyMo and
Track mechanism the ground in order to minimize the torque required to climb the step. The simulations
with different contact angles are carried out and then, the experiments using the RHyMo
are performed, which can demonstrate that the trend of measured motor current is similar
to that simulated by the kineto-static analysis and moreover, the step-climbing capability
of the RHyMo is significantly improved by reducing the required motor currents by virtue
of the kineto-static analysis.
© 2017 Elsevier Ltd. All rights reserved.

1. Introduction

As the robots have been widely adopted in a variety of fields with the advance of related technologies, there has been an
increasing demand for the mobile robots which can complete their tasks irrespective of where they are. Therefore, it is not
surprising that the mobility becomes the most essential factor when the performance of mobile robot is evaluated [1–6].
Since the performance of mobile robot may be significantly influenced by the shape of ground as well as the unexpected
impact from the ground which the mobile robot keeps in contact with, it is highly required to control the mobile robot
appropriately based on the dynamic analysis including the force and moment relations between the mobile robot and the
ground [7,8].
Among various metrics to evaluate the performance of mobile robot, the required friction coefficient has been widely
adopted, which is defined as the minimum friction coefficient required for a mobile robot to climb a step without slip [1].
Since the height of step is set equal to the wheel diameter of mobile robot when calculating the required friction coefficient,


Corresponding authors.
E-mail addresses: hskim94@kgu.ac.kr (H.S. Kim), jongkim@snu.ac.kr (J. Kim).
1
Both of these authors contributed equally to the work in this paper.

http://dx.doi.org/10.1016/j.mechmachtheory.2017.03.018
0094-114X/© 2017 Elsevier Ltd. All rights reserved.
D. Choi et al. / Mechanism and Machine Theory 114 (2017) 20–37 21

the size of mobile robot has no effect on this metric. It is worthwhile to note that in order to calculate the required friction
coefficient, the normal/traction forces acting on each wheel of mobile robot are necessary. For a long time, the quasi-static
analysis has been used due to its simplicity and reasonable accuracy to calculate the normal/traction forces and torques
acting on the mobile robot. Under the quasi-static condition, the mobile robot is assumed to move at a sufficiently low and
constant speed as if its movement consists of consecutive static equilibriums [9–12]. For examples, the quasi-static analysis
can be adopted to analyze the behavior of Mars Exploration Rovers (MERs) which travel on rugged terrain at low speeds of
3–5 m/min [13,14]. However, for mobile robots whose moving speeds are maximally up to a few dozen meters per minute,
their dynamic effects cannot be neglected in order to capture the characteristics of their dynamic behavior quickly and
precisely so that the quasi-static analysis does not seem suitable for the force/torque analysis of high-speed mobile robots
[15,16]. Recently, Jun et al. evaluated the performance of leg-wheel locomotion subsystem via the kineto-static analysis
considering the kinematic link length as well as the static spring stiffness and preload [17]. Alamdari et al. performed the
kineto-static optimization to determine the spring stiffness of four-bar based articulated leg-wheel subsystem [18]. Also,
Alamdari et al. developed the kineto-static based optimization for design of articulated leg-wheel subsystem with the aim
of maximizing the motion-ranges and reducing the actuation requirements [19]. These works have considered the kinematics
as well as the statics for optimal design of leg-wheel subsystem but due to the complexity of trajectory of terrain, the effect
of dynamics has not been fully reflected.
In order to consider the effect of dynamics such as the speed and acceleration of center of mass (CM) of mobile robot
along the moving direction, Zvi Shillar et al. suggested a new dynamic analysis method which can determine the feasible
ranges of speed/acceleration of CM of mobile robot to satisfy a set of dynamic constraints, for examples, normal force on
a wheel must be positive for a mobile robot to keep in contact with the ground and the ratio of traction force to normal
force must be smaller than friction coefficient of ground for a mobile robot to move without slip [20,21]. Therefore, this
analysis method may help to conclude whether the mobile robot can travel on the terrain stably and quickly with the given
speeds and accelerations but cannot determine the forces and torques acting on the mobile robot uniquely. Of course, the
normal/traction forces and torques may be obtained by using the pseudo-inverse matrix but in some cases, the computed
forces or torques may be negative to violate the assumption that all wheels of mobile robot keep in contact with the ground
or go forward [22].
This study first presents the kineto-static analysis for the relatively high-speed mobile robot while climbing a step. In
comparison with the existing kineto-static analysis, the proposed method reflects the dynamic effects of mobile robot such
as the speed and acceleration while the mobile robot climbs a step. To this end, the kinematic analysis is carried out to
calculate the trajectory of center of mass (CM) of mobile robot during climbing a step. Based on the calculated CM trajectory
of mobile robot, the overall trace of each link can be obtained. Then, by selecting two points on the resulting trace of the
link and dividing their linear/angular differences with the constant time step, the linear/angular speeds of the links can
be calculated. To calculate the linear/angular accelerations of links, the same procedure can be applied to three points on
the overall trace of link. The dynamic equations of mobile robot are constructed by considering the force and moment
equilibriums of mobile robot on the terrain. By combining the speed and acceleration data obtained from the kinematic
analysis with the dynamic equations of mobile robot, it is possible to quickly capture the key characteristics of dynamic
behavior of mobile robot climbing the step. However, it should be noted that the dynamic equations of wheeled mobile
robot interacting with the environment are likely to be indeterminate due to too many unknown parameters such as a
traction/normal forces and friction coefficients. Therefore, additional constraints are required to solve these indeterminate
dynamic equations. For examples, in the case of the Rocker–Bogie mechanism, the normal forces acting on all wheels may
be assumed to be equal because of its kinematic structure but recall that this assumption is valid only for the case that
the Rocker–Bogie mechanism travels on a flat surface [1,3]. The stair-climbing wheeled mobile robot (WMR) developed by
Woo et al was designed to have the maximum traction forces of some wheels on a flat surface via the optimization of its
structure [10]. However, this method was performed by assuming the quasi-static condition and still has the drawback that
other traction forces of wheels on a vertical surface cannot be calculated. Therefore, in order to solve the indeterminate
dynamic equations of the mobile robot, this study adopts the concept of required friction coefficient that corresponds to
the minimum friction coefficient required for the mobile robot to maintain the equilibrium while overcoming an obstacle
without slip.
The proposed kineto-static analysis is applied to the mobile robot RHyMo shown in Fig. 1, which is developed by com-
bining the Rocker–Bogie mechanism with the inverse four bar mechanism in order to ensure the smooth movement on
rugged terrains [23]. First, assuming that the RHyMo climbs the step at the speed of 10 m/min, the kineto-static analysis is
carried out to calculate the normal/traction forces and torques acting on the links and wheels of RHyMo by minimizing the
required friction coefficient and then, the obtained normal/traction forces and torques by the proposed kineto-static method
are compared to those obtained by the quasi-static analysis. Subsequently, the kineto-static analysis based optimization is
repeatedly carried out, where the contact angle between the front track of RHyMo and the ground is changed in order to
find out the contact angle which can reduce the peak value of the maximum torque required for the RHyMo to climb the
step. The experiments using the RHyMo with various contact angles are performed, which prove that the trends of measured
motor currents are very similar to those of motor torques obtained by the kineto-static analysis and also that the required
motor current is considerably reduced by properly adjusting the contact angle of front track of RHyMo. It is worthwhile to
note that the main contribution in [23] is to develop a new mobile platform to ensure smooth movement on rugged terrains.
To this end, the new criterion called as the posture variation index (PVI) was suggested to evaluate the movements of var-
22 D. Choi et al. / Mechanism and Machine Theory 114 (2017) 20–37

Fig. 1. Photograph of mobile robot RHyMo equipped with small robots and quadcopter.

ious wheeled and tracked mobile platforms systematically. Then, the new mobile platform RHyMo was optimally designed
to minimize this PVI through the kinematic and quasi-static analysis. The experiments using the artificial terrain and stairs
proved that the movement of RHyMo was much smoother than that of well-known Rocker–Bogie platform. On the contrary,
the contribution in this paper is to reduce the motor torque for the rear wheel of the RHyMo required to climb a step.
To this end, the kineto-static analysis is repeatedly performed by changing the contact angle of first track of RHyMo. The
extensive experiments are carried out to measure the motor current of rear wheel of RHyMo during climbing the 150 mm
high step, whose trend seems similar to the torque trend of rear wheel of RHyMo obtained by the kineto-static analysis and
moreover, the motor current of rear wheel of RHyMo can be significantly reduced by adjusting the contact angle of front
track of RHyMo properly.
The rest of this paper is organized as follows. In Section 2, the trajectory, speed and acceleration data of CM of the
RHyMo during climbing the step are obtained through the kinematic analysis. In Section 3, the kineto-static analysis of the
RHyMo is described to compute the normal/ traction forces and torques acting on its wheels by using the concept of the
required friction coefficient. In Section 4, the experimental results using the RHyMo for the 150 mm high step are given to
prove the viability of the proposed kineto-static analysis.

2. Kinematic analysis of RHyMo

2.1. Generation of CM trajectory

It is well known that through the dynamic analysis, the next position and posture of mobile robot can be obtained from
the information of the current position, posture, velocity and acceleration as well as the forces and moments acting on the
mobile robot. To the contrary, the inverse of dynamic analysis is to calculate the forces and moments acting on the mobile
robot by using a prior knowledge of its entire positions, postures, velocities and accelerations [24–26]. Therefore, the known
position, postures, velocities and accelerations of mobile robot are essential for the inverse of dynamic analysis. Please recall
that the main thrust of this study is to enhance the climbing capability of RHyMo against the step by reducing the torque
required for its rear wheel. Therefore, we have focused on capturing the key characteristics of dynamic behavior of RHyMo
during climbing a step rather than making the dynamic model for the RHyMo. In fact, the accelerometer is known to be
quite vulnerable to the noises or oscillations that the RHyMo may undergo during climbing a step. From this viewpoint,
this study adopts the kineto-static analysis using the positions/postures, velocities and accelerations obtained through the
kinematic analysis. Note that it is quite difficult to apply the kineto-static analysis to the mobile robot traveling on the
rugged terrain because it is almost impossible to obtain the precise profile of rugged terrain in real time. In this study, the
positions and postures of RHyMo climbing a 150 mm high step are first calculated through the kinematic analysis. Recall
that the RHyMo is assumed to keep in contact with the ground while climbing the step so that its positions and postures
can be uniquely determined. Also, for the simplicity of analysis, the CM trajectory of RHyMo is considered in 2-dimensions.
Fig. 2(a) and (b) show the procedure to calculate the CM trajectory of RHyMo climbing the step and how to choose the
reference points for the RHyMo according to the contact condition between the track and the ground. It is assumed that
the moving speed of RHyMo is 10 m/min so that the distance between reference points is chosen to be 0.835 mm with the
time interval of 0.005 s.
D. Choi et al. / Mechanism and Machine Theory 114 (2017) 20–37 23

Fig. 2. (a) Procedure of calculating the CM trajectory of RHyMo climbing a step and (b) schematic diagram of defining the reference points for RHyMo
according to the contact conditions.

According to the contact condition between the track and the ground, the reference points for the RHyMo may be chosen
as follow: if l1 sin θ0 − r1 cos θ0 + r1 < h holds, the front wheel of track comes in contact with the riser of step as shown in
(case 1) of Fig. 2(b) and then, the reference points are located on the side of track, the flat surface and the riser of step.
To the contrary, if l1 sin θ0 − r1 cos θ0 + r1 ≥ h is satisfied, the side of track directly makes in contact with the corner of step
as shown in (case 2) of Fig. 2(b) and then, the reference points are located on the side of track and flat surfaces of step.
After the track of RHyMo is positioned on the certain reference point, the resulting positions of middle and rear wheels are
sequentially determined by the geometric conditions on the RHyMo, that is, the initial value of θ 1 is set as 90° and then,
increased in a counter-clockwise direction by the variation θ 1 until the middle wheel comes in contact with the ground.
The same method is applied to the rear wheel. After determining the positions of track and two wheels, the resulting
posture and position of links are computed via the kinematic analysis. This procedure is repeated by moving the RHyMo
along with the reference points until the track arrives at the final destination.

2.2. Kinematic model for RHyMo

It is worthwhile to note that when the track and two wheels of RHyMo are suitably located on the step, the resulting
positions and postures of its links can be determined via the kinematic analysis. Fig. 3 shows the schematic diagram of
24 D. Choi et al. / Mechanism and Machine Theory 114 (2017) 20–37

Fig. 3. Schematic diagram of RHyMo for kinematic analysis.

RHyMo for the kinematic analysis, where A, B, C denote the centers of track and two wheels and D, F, H and I represent
the revolute passive joints and, E denotes the CM of main body of RHyMo, respectively. The lengths of links AD, BD, DF, DI,
CI, HI and DI are denoted by s1 , s2 , s9 , s10 , s11 , s12 and s13 , and the distances between A and B, D and E are given by s0 and
Ls , respectively. The contact angle between the track or wheels and the ground are denoted by α i and the angles of Bogie
and Rocker mechanisms are given by θ i , i = 1, 2, 3, respectively. The angles between links AD and BD, BD and DF, CI and
HI are denoted by ϕ i , i = 1, 2, 3, respectively. Note that the lengths of links (s1 , s2 , s9 , s10 , s11 , s12 , s13 ) and the angles ϕ i are
the design parameters and the conditions (α i , hi ), i = 1, 2, 3 are given by the surface of ground that the RHyMo travels on.
Based on the prior information of these parameters and conditions, the angles between the link and the ground θ k , k = 1,
, 5, can be calculated.
From the geometric condition on Bogie mechanism in Fig. 3, θ 1 can be obtained by
   
r2 cos α2 + h2 − r1 cos α1 − h1 s2 sin ϕ1
θ1 = sin−1 − tan−1 (1)
(s2 cos ϕ1 )2 + (s1 − s2 cos ϕ1 )2 s1 − s2 cos ϕ1
where ϕ1 = cos−1 (s21 + s22 − s20 /2s1 s2 ). Then, θ 2 can be easily calculated by using θ2 = θ1 − ϕ1 . By considering two closed-
loops DFHID and BCIDB, the following relations are derived:

s10 sin θ4 = c1 
+ s11 sin θ3
(2)
s10 cos θ4 = − s210 − (c1 + s11 sin θ3 )
2

 
s13 sin θ5 = c32 + c42 sin(θ3 + φ ) + c1 + c2
  (3)
2
s13 cos θ5 = − s213 − ( c32 + c42 sin(θ3 + φ ) + c1 + c2 )
 
 2
s210 − (c1 + s11 sin θ3 ) + s12 cos(ϕ3 − θ3 ) = s213 − ( c32 + c42 sin(θ3 + φ ) + c1 + c2 ) + c5
2
(4)

where c1 = h3 + r3 cos α3 − h2 − r2 cos α2 − s2 sin θ2 , c2 = s9 sin(θ2 − ϕ2 ), c3 = s11 − s12 cos ϕ3 , c4 = s12 sin ϕ3 , c5 = s9 cos(θ2 −
ϕ2 ) and φ = tan−1 (c4 /c3 ). Please, refer to [23] for more detailed explanation of deriving the kinematic model of RHyMo.
Note that if θ 3 is calculated from Eq. (4), θ 4 and θ 5 can be obtained from Eqs. (2) and (3). However, since Eqs. (2)–(4)
consist of cosine and sine functions, these equations are highly nonlinear so that in this study, the exhaustive search method
is used to precisely calculate a value for θ 3 .
Table 1 presents the detailed specifications of links, tracks and wheels of RHyMo. It is worthwhile to note that the
lengths of links are optimally chosen in such a manner to reduce the height and pitch angle variations of the CM when the
RHyMo travels on rugged terrain. To this end, the posture variation index (PVI) is defined, which corresponds to combined
variations in height as well as pitch angle of CM of mobile platform when its wheel is lifted up to the height of its wheel
radius as given in Eq. (5) [23]:

N
N
Hmobile Nmobile
PV Imobile = × (5)
H42W D 24W D
 
avg 2 2 2 avg 2 2 2
where N
Hmobile = a(Hmobile ) + b(Hmobile
dev ) + c (H max )
mobile
and Nmobile = a(mobile ) + b(de v ) + c (max ) .
mobile mobile
avg avg
Hmobile (mobile ),
dev (dev ) and H max (max ) are the average height (pitch angle) variation, its standard devia-
Hmobile mobile mobile mobile
tion and the maximum height (pitch angle) variation, respectively. The weighting factors a, b and c are adopted to
D. Choi et al. / Mechanism and Machine Theory 114 (2017) 20–37 25

Table 1
Specifications of RHyMo.

Part Length (m) Weight (kg) Moment of inertia (kg m2 )

Link0 s0 0.300 5.870 0.138


s1 0.180
s2 0.260
s9 0.084
Link10 s10 0.318 1.305 0.031
Ls 0.209
Link11 s11 0.241 3.390 0.045
s12 0.066
Link13 s13 0.272 1.415 0.019
Track l1 0.220 4.065 0.023
wheel ri 0.093 0.680 0.003
Main body 1 × 0.7 × 0.45 (L × W × H) 25.3 0.943

reflect the effect of each term in Eq. (5). H42W D and 24W D correspond to the total height and pitch angle variations of a
four-wheeled mobile platform of same dimensions which has no suspension system. We first evaluated the motions of
various well known mobile platforms by using this PVI. For the simplicity of PVI evaluation, the height and pitch angle
variations of CM of wheeled mobile platform are measured when the ith wheel is sequentially lifted up to its wheel radius
above the ground. The PVI evaluation on various mobile platforms revealed that the movement of Rocker–Bogie platform
was much smoother than other platforms, for examples, SHRIMP, RCL-E, CRAB and PackBot, etc. It is noted that the PVI
of Rocker–Bogie is drastically deteriorated for the rear wheel because the CM of Rocker–Bogie lies on the Rocker link that
the rear wheel is directly connected to so that it is important to remove the correlation between the rear wheel and the
CM by using additional mechanism. Therefore, in order to improve the smoothness of Rocker–Bogie’s movement on rugged
terrains, we suggested a new mobile platform combining the Rocker–Bogie with an inverse four bar mechanism. Since the
motions of links of four bar mechanism can be predicted for the given link lengths, it is possible to tune the link lengths
and angles of RHyMo by the exhaustive search method so as to minimize the resulting PVI. For more detailed information,
please refer to [23]. The moment of inertia of each link is computed along the Z-axis in Fig. 3. By using the specifications
of the RHyMo as well as the prior information of ground condition (α i , hi ), i = 1, 2, 3 given by the step, the angles between
the link and the ground θ k , k = 1, , 5 can be obtained through the kinematic analysis at every reference point.
The resulting CM trajectory of the RHyMo climbing a 150 mm high step is shown in Fig. 4(a), where the red circles de-
note the moments when the track and wheels of RHyMo climb the step and the CM trajectory is denoted by the green line.
In comparison with the CM trajectory of Rocker–Bogie mechanism suffering from undesired backward movement [12], the
CM trajectory of RHyMo seems much smoother. And it is drastically changed three times along the Y-axis when the track
and wheels of RHyMo come in contact with the step and climb up it as shown in Fig. 4(a). Note that when the track or
wheel climbs up the step, the position/posture of each link of RHyMo must be adjusted and as a result, the linear/angular
accelerations are naturally generated by the interaction between the ground and the track/wheels. Recall that the difference
between two positions/angles on the CM trajectory of RHyMo is used to calculate linear/angular speeds of RHyMo and, the
difference between two linear/angular speeds on the CM trajectory of RHyMo is used to calculate linear/angular acceler-
ations of RHyMo. Fig. 4(b) compares the real and simulated CM trajectories, where the real trajectory of CM seems quite
similar to the trajectory obtained by the kinematic analysis. As indicated by red dotted line in Fig. 4(b), a small discrepancy
between real and simulated trajectories of CM is observed, which can be explained by the fact that in real experiment, the
contact angle of front track is increased due to the backlash of front track motor and as a result, the real CM of RHyMo
moves ahead of simulated CM. Also, the elastic deformation of real wheels and the irregular thread of wheel or track may
have the effects on the real trajectory of CM.
The linear/angular speeds of CM of RHyMo are described in Fig. 5, where the orange solid, blue dotted and magenta
dash-dotted lines denote the linear speeds along the X- and Y- axes, and the angular speed along the Z-axis, respectively.
The moving speed of RHyMo is 10 m/min (1.66 m/s) so that when the RHyMo travels on a flat surface, its speeds along the
X- and Y-axes are 1.66 m/s and zero, respectively. Compared to the cases that the middle and rear wheels climb the step,
the speed of CM along the X-axis is significantly changed so that it is reduced to almost zero even before the track starts to
climb the step. It is worthwhile to note that when the middle and rear wheels climb the step, the resulting speed changes
along the Y-axis are relatively larger than the speed change along the Y-axis when the front track climbs the step, which
implies that the front track may reduce the additional force required to lift up the RHyMo along the Y-axis in comparison
with the wheel. In fact, the force required to move the front track along the X-axis is relatively larger than the force required
to lift up it along the Y-axis. This observation is confirmed in Fig. 6 which describes the linear/angular accelerations along
the X-, Y- and Z-axes. As shown in Fig. 6, the maximum accelerations along the X- and Y-axes occur when the front track
and the middle wheel start to climb the step, respectively. To the contrary, the maximum angular acceleration along the
Z-axis occurs when the rear wheel starts to climb the step. It is noted that the linear/angular accelerations in Fig. 6 can be
used to calculate the normal/traction forces and torques in combination with the force/moment equilibriums of RHyMo.
26 D. Choi et al. / Mechanism and Machine Theory 114 (2017) 20–37

Fig. 4. (a) Simulated CM trajectory of RHyMo while climbing a 150 mm high step and (b) comparison of simulated and real CM trajectories.

Fig. 5. Linear/angular speeds of CM of RHyMo while climbing a 150 mm high step.

3. Kineto-static analysis of RHyMo

3.1. Force and moment equilibrium equations for RHyMo

The required friction coefficient is known as the reasonable metric to evaluate the terrainability of mobile robot, which
is defined as the minimum friction coefficient required for a mobile robot to climb a step whose height is equal to its wheel
D. Choi et al. / Mechanism and Machine Theory 114 (2017) 20–37 27

Fig. 6. Linear/angular accelerations of CM of RHyMo while climbing a 150 mm high step.

Fig. 7. Free body diagram of RHyMo for kineto-static analysis.

diameter [1]. To calculate the required friction coefficient, the normal/traction forces acting on each wheel are indispensable.
In this section, the kineto-static relations for the RHyMo will be established and solved by the concept of required friction
coefficient using the acceleration data obtained by the kinematic analysis in Section 2.
Fig. 7 shows the free body diagram of RHyMo for kineto-static analysis, where the traction and normal forces and torques
of track and wheels are represented by Ti , Ni and τ i , i = 1, 2, 3, respectively. It is worthwhile to note that the deep groove
ball bearings are used for all active and passive joints denoted by A, B, C, D, F, H and I in Fig. 3 so that the effects of
frictions at these joints are negligible during the kineto-static analysis in comparison with the traction and normal forces
acting on the track and wheels of RHyMo because the friction coefficient of deep groove ball bearing is typically less than
F
0.002 [27]. mi and i k denote the mass of ith element and the shortest distance vector between the force Fk and the CM of
ith element, respectively. aix (aiy ) and ω˙ i denote the linear and angular accelerations of ith element along the X-axis (Y-axis)
and the Z-axis, respectively. ri and Ii denote the radius and the moment of inertia of ith element with respect to the Z-axis,
respectively.
Under the conditions on the RHyMo in Fig. 7, the force and moment equations related with the front track along the X-,
Y- and Z-axes can be obtained as follow:

−F1x + T1 cos α1 − N1 sin α1 = m1 a1x (6)

−F1y + T1 sin α1 + N1 cos α1 = m1 a1y + m1 g (7)

−τ1 + r1 T1 − m1 g · m
1
1g
= I1 ω˙ 1 (8)
28 D. Choi et al. / Mechanism and Machine Theory 114 (2017) 20–37

By using T1 = μ1 N1 and T2 = μ2 N2 , the force and moment equations related with the middle and rear wheels along the
X-, Y- and Z-axes are given by
−F2x + (μ2 cos α2 − sin α2 )N2 = m2 a2x (9)

−F2y + (μ2 sin α2 + cos α2 )N2 = m2 a2y + m2 g (10)

−τ2 + μ2 r2 N2 = I2 ω˙ 2 (11)

−F3x + (μ3 cos α3 − sin α3 )N3 = m3 a3x (12)

−F3y + (μ3 sin α3 + cos α3 )N3 = m3 a3y + m3 g (13)

−τ3 + μ3 r3 N3 = I3 ω˙ 3 (14)
where g is the acceleration of gravity. For the 4th element including the Bogie mechanism denoted by the Link0 in Fig. 3,
the force and moment equations along the X-, Y- and Z-axes are given by
F1x + F2x + FDx + FFx = m4 a4x (15)

F1y + F2y + FDy + FFy − m4 g = m4 a4y (16)

F F F F
τ1 + τ2 + F1x F41x + F1y 41y + F2x F42x − F2y 42y − FDx F4Dx + FDy 4Dy − FFx F4Fx − FFy 4Fy = I4 ω˙ 4 (17)
For the 5th element including the CM denoted by the Link10 in Fig. 3, the force and moment equations along the X-, Y-
and Z-axes are given by
−FDx + FIx = m5 a5x + Mb a5x (18)

−FDy + FIy − m5 g − Mb g = m5 a5y + Mb a5y (19)

F FD F FI FM
FDx 5Dx − FDy 5 y + FIx 5Ix − FIy 5y − Mb g5 bg
= I5 ω˙ 5 + IMb ω˙ 5 (20)
For the 6th element connecting the Bogie and the Rocker mechanisms denoted by the Link13 in Fig. 3, the force and
moment equations along the X-, Y- and Z-axes are derived as follow:
−FFx + FHx = m6 a6x (21)

−FFy + FHy − m6 g = m6 a6y (22)

−FFx 6FF − FFy 6FF − FHx 6FH − FHy 6FH = I6 ω˙ 6 (23)


x y x y

For the 7th element including the Rocker mechanism denoted by the Link11 in Fig. 3, the force and moment equations
along the X-, Y- and Z-axes are derived as follow:
F3x − FIx − FHx = m7 a7x (24)

F3y − FIy − FHy − m7 g = m7 a7y (25)

F F F
τ3 + F3x F73x − F3y 73y + FIx F7Ix − FIy 7Iy + FHx F7Hx − FHy 7Hy = I7 ω˙ 7 (26)
It is worthwhile to note that the total number of unknown variables is 23 because there are 6 reaction forces (Fix and Fiy ,
i = 1, 2, 3) of active joints denoted by A, B and C in Fig. 3, 8 reaction forces (Fjx and Fjy , j=D, F, I, H) of revolute joints denoted
by D, F, H and I in Fig. 3, 6 external forces (Ni and Ti , i = 1, 2, 3) acting on the track and wheels and 3 torques (τ i ,i = 1, 2,
3) for the track and wheels. Recall that to solve the problem under the kineto-static condition is indeterminate because the
number of constraints from Eqs. (6) to (26) is smaller than the number of unknown variables. In order to tackle this problem
more effectively, the constraints from Eqs. (6) to (26) are recast into a simple algebraic form with less unknown variables.
In this study, the normal forces N1 , N2 , N3 , the traction force T1 , the reaction forces FIx , FIy and the friction coefficients μ2 ,
D. Choi et al. / Mechanism and Machine Theory 114 (2017) 20–37 29

Fig. 8. Normal/traction forces and motor torques for front track, middle/rear wheels of RHyMo to climb the 150 mm high step.

μ3 are chosen as the unknown variables to be solved first. After some algebraic manipulation of Eqs. (6)–(26), the following
matrix equation can be derived (for more details, please refer to the Appendix):
A6×6 · X6×1 = B6×1 (27)
where X6×1 = [N1 N2 N3 T1 FIx FIy ]T .
In order to simplify this indeterminate problem in Eq. (27), without loss of generality, it
is assumed that the friction coefficients in the matrix A6 × 6 range from 0.1 to 2.0 with the step size of 0.01. Then, by simply
using the inverse matrix ofA6 × 6 , the unknown variables X6 × 1 can be obtained by X6×1 = A−1 6×6
· B6×1 for the given friction
coefficients. Since under the assumption of range of friction coefficient, the total number of possible solutions to Eq. (27) is
191 × 191 so that the optimization method is naturally adopted to impose additional constraint on Eq. (27). So, this study
utilizes the concept of the required friction coefficient which corresponds to the minimum friction coefficient required for
the mobile robot to overcome the step without slip. Since the friction coefficient can be calculated by dividing the traction
force by the normal force [1] and the RHyMo is supposed to keep a good contact with the ground, the optimal values for the
unknown variables can be obtained by solving the following the optimization problem subject to constraint in Eq. (27) and
the assumption on range of the friction coefficient as well as good contact with the ground:
 F F2 F3
1
min max , , (28)
N1 N2 N3
subject to A6×6 · X6×1 = B6×1 , 0.1 ≤ μi ≤ 2 (μi = 0.01) and Ni > 0, i = 1, 2, 3.
Fig. 8 shows the normal/traction forces and motor torques acting on the front track and middle/rear wheels obtained by
solving Eq. (28) under the assumptions on friction coefficient as well as good contact with the ground. In Fig. 8, the regions
painted by the blue, red and green colors denote the moments that the front track, the middle and rear wheels of RHyMo
climb up the step, respectively. According to Fig. 8, the normal force for the front track of RHyMo is larger than those
for the middle/rear wheels so that from the viewpoint of the force distribution, the normal force of RHyMo is relatively
concentrated on the front track. Therefore, if the front link of RHyMo is equipped with a wheel, the resulting climbing
capability of RHyMo may be deteriorated because the traction force of wheel is obviously smaller than that of the track.
Unlike the normal force, the trends of traction force and motor torque are very similar to each other. As confirmed in Fig. 8,
the maximum motor torque occurs at the rear motor when the front track climbs the step after contacting with it. It is
worthwhile to note that the maximum motor torque can be changed according to the posture of mobile robot so that it is
30 D. Choi et al. / Mechanism and Machine Theory 114 (2017) 20–37

Fig. 9. Comparison of torques for (a) front track, (b) middle wheel and (c) rear wheel obtained by the quasi-static analysis and kineto-static analysis.

necessary to perform the kineto-static analysis based optimization with various postures in order to reduce the maximum
motor torque while climbing the step, which will be in details discussed in the next section.

3.2. Investigation of effect of contact angle of RHyMo via kineto-static analysis

The proposed kineto-static analysis is compared to the existing quasi-static analysis in this section. Fig. 9 describes the
torques for the front track and the middle/rear wheels of RHyMo obtained by the quasi-static analysis and the kineto-static
analysis. Note that under the quasi-static condition, it is sufficiently enough to maintain the mobile robot statically while
climbing a step, which implies that the traction and normal forces acting on the track or wheels have only to support the
D. Choi et al. / Mechanism and Machine Theory 114 (2017) 20–37 31

Fig. 10. Torque for the rear wheel of RHyMo according to the contact angle of front track.

mass of mobile robot while climbing the step. However, since the moving speed as well as the acceleration of RHyMo are
considered under the proposed kineto-static analysis, the required torques for the track and wheels may be influenced in a
positive or negative way.
In Fig. 9, the regions painted by the blue, red and green colors denote the moments that the front track, the middle and
rear wheels climb up the step, respectively. For an example, the red region in Fig. 9(a) shows the required torque for the
front track when the middle wheel of RHyMo climb up the step. In comparison with the required torques obtained by the
quasi-static condition, the required torques for the front track and middle wheel are increased as shown in Fig. 9(a) and (b)
but the required torque for the rear wheel is decreased as shown in Fig. 9(c) when the front track and middle wheel are
already on the step. So, when the front track climbs the step, the required torques for the front track and middle wheel are
increased by 9% and 9.5%, but the torque for the rear wheel is decreased by 7.8% as confirmed in the blue region in Fig. 9.
Similarly, when the middle wheel climb the step, the required torques for the front track and middle wheel are increased by
10.7% and 10.9%, but the required torque for the rear wheel is decreased by 4.3% as confirmed in the red region in Fig. 9. It
is worthwhile to note that when the front track or the middle wheel climbs the step, the joint D drastically goes up to make
the Link10 rotate in the counter-clockwise direction, which moves the CM of RHyMo backward and as a result, additional
torque is required for the RHyMo to go forward with maintaining its moving speed. On the contrary, when the rear wheel
climbs the step, the Link10 rotates in the clockwise direction and as a result, the CM of RHyMo is naturally accelerated
along the moving direction. Therefore, in this case, the torque obtained by the proposed kineto-static analysis becomes
smaller than the torque obtained by the quasi-static condition. As shown in the green region in Fig. 9, when the rear wheel
climb the step, all torques for the front track, middle and rear wheels are decreased by 6.6%, 13.9% and 20.4%, respectively. In
comparison with the quasi-static analysis, the kineto-static analysis can reflect the dynamic effects of linear/angular speeds
and accelerations of RHyMo so that it is possible to predict the key characteristics of dynamic behavior of RHyMo more
accurately.
It is noted that the RHyMo is designed to adjust the contact angle between the front track and the ground from 0° to
90°. Recall that when the track of RHyMo climbs up the step, the maximum torque occurs at the rear wheel of RHyMo as
shown in the blue region in Fig. 9(c). To find out the optimal contact angle between the front track and the ground during
climbing the step, the extensive simulations using the proposed kineto-static analysis are carried out with various contact
angles. Note that while the five simulations set the constant contact angles to be 15°, 25°, , 65°, the contact angle in the
6th simulation is changed from 45° to 5° during climbing the step. The resulting torques for the rear wheel are shown in
Fig. 10. When the range of contact angle is from 15° to 35°, the torque is increased at the moment that the track starts to
contact with the step and then, is decreased while the track climbs up the riser of step. However, when the track reaches
the corner of step and the side of track moves along the corner, the torque is increased again so that the shape of torque is
similar to the small letter ‘m’.
On the other hand, when the contact angle ranges from 45° to 65°, the shape of torque is similar to the small letter
‘n’ since the side of front track directly comes in contact with the corner of step so that no sharp peak in the torque is
observed compared to the case of the contact angle ranging from 15° to 35°. The peak value of torque for the rear wheel
occurs when the bottom of front track stands upon the step. As the contact angle between the front track and the ground
is increased, the peak value of torque for the rear wheel is also increased. According to the simulations, when the contact
angle is fixed, the smallest peak value of torque for the rear wheel is achieved (=about 3.5 Nm) at the contact angle of 35°.
32 D. Choi et al. / Mechanism and Machine Theory 114 (2017) 20–37

It should be noted that when the contact angle of track is smaller, the torque required for the rear wheel is increased in
order for the track to climb the riser of step without losing contact with the step. To the contrary, when the side of track
climbs up the corner of step, the required torque is decreased as the contact angle of track is smaller. Therefore, in order
for the RHyMo to climb the step with much smaller torque, the contact angle of track must be set to be large enough to
make the side of track contact with the corner of step directly and then, adjusted to be a smaller one. As shown in Fig. 10,
when the contact angle of track is changed from 45° to 5° during climbing the step, the peak value of torque for the rear
wheel is significantly reduced from 3.5 Nm to 1.5 Nm in comparison with the torque obtained when the contact angle is
set to be 35°. Even though a mathematical model-based optimization is not fully carried out in this section, please note
that the kineto-static simulation in Fig. 10 is basically constructed on the optimization in Eq. (28) and also, the proposed
kineto-static analysis seems to play a sufficient role in delivering the key clue for minimizing the peak value of torque for
rear wheel of RHyMo, which will be verified in the next section through the extensive experiments.

4. Experimental validation and discussions

4.1. Mobile robot RHyMo

The RHyMo is the mobile robot whose linkage mechanism is based upon the combination of the Rocker Bogie, the
inverse four bar mechanism and the front track with the aim of reducing undesired variations in height and pitch angle
while traveling on rugged terrain. The overall size of RHyMo is 700 mm x 450 mm x 1000 mm (width x height x length) and
its weight and maximum payload are 53 kg and 60 kg, respectively. The proposed RHyMo is controlled in real time by a
human using a Wii Bluetooth controller that gives commands to the motor drives via wireless connection with the Arduino
module. As shown in Fig. 1, the motors for driving the track and wheels are placed inside the hollow links or parallel to
the links to avoid interference between links or link and the ground. Also, additional motor is located inside the front link
to control the contact angle of front track from 0° to 90°. The maximum moving speed of RHyMo on a flat surface is up
to 20 m/min and the maximum torque of track or wheel is up to 15 Nm which is three times larger than the required
maximum torque obtained in the kineto-static analysis.

4.2. Experiment using a step without adjusting contact angle

The experiment using the step is performed in order to investigate the viability of proposed kineto-static analysis. The
moving speed of RHyMo is 10 m/min and the height of single step is set to be 150 mm which is similar to the wheel
diameter. The contact angle between the track and the ground is chosen to be 38.5°. According to the simulation results in
the previous section, the trend of torque for the rear wheel is expected to be similar to the alphabet ‘N’ while the front track
climb the step and its maximum will occur when the end of track climbs the corner of step. Fig. 11 shows the snapshots
of RHyMo’s climbing the step, where the side of front track directly comes in contact with the step and climbs the step
without changing the contact angle of front track. It should be noted that when the front track of RHyMo in Fig. 11 is
replaced with the wheel, the RHyMo cannot overcome the step in the real experiment, which is because the maximum
friction coefficient of RHyMo is increased by using the wheel and also, the contact point between the wheel and the step
becomes much lower [23].
It is worthwhile to note that if the response of motor is quite fast, the resulting motor torque can be assumed to be
in proportion to the motor current so that in this study, without loss of generality, the motor torque is approximated by
the product of the motor current and the torque constant [28,29]. The motor current for the rear wheel is measured by
the current sensor in the motor drive and its trend is compared to that of torque obtained by the kineto-static analysis
in Fig. 10. The measured motor current is shown in Fig. 12(a), where the orange solid and blue dotted lines denote the
simulated motor torque for the rear wheel and the motor currents for the rear wheel, respectively. As shown in Fig. 12(a),
the overall trend of motor current obtained by the experiment is quite similar to that of motor torque obtained by the
simulation so that the maximum motor current occurs when the bottom of track reaches the corner of step. However, it
is worthwhile to note that during the RHyMo’s traveling on a flat surface, the trend of motor current is slightly different
from that of motor torque as if there seems a certain offset. In the simulation, if the moving speed of RHyMo is constant,
the required torque becomes zero since additional acceleration is not necessary. Recall that the non-zero minimum power
is required to not only rotate the motor even on a flat surface but also overcome the friction forces in the experiment.
It is worthwhile to note that when establishing the force and moment equilibrium equations for the RHyMo, it is as-
sumed that there is no slip between wheels and the ground by introducing the concept of required friction coefficient.
Also, the uniform contact between the track or wheels and the ground is assumed. However, as seen in the attached video
files, slight slip is observed when the front track or wheels climb up the step and the irregular contact between the track or
wheels and the ground sometimes occurs due to the threads of wheels and track, which may lead to undesired oscillation of
the traction force acting on the track or wheels. Even though considering different physical units are used, there is naturally
a discrepancy between the measured motor currents and the simulated motor torque for rear wheel of RHyMo as shown
in Fig. 12(a). On the other hand, in spite of such a discrepancy, the trend of measured motor current is very similar to that
of simulated motor torque, which implies that the simulated torque obtained by the kineto-static analysis plays a sufficient
D. Choi et al. / Mechanism and Machine Theory 114 (2017) 20–37 33

Fig. 11. Snapshots of RHyMo’s climbing the step with the constant contact angle of 38.5°

Fig. 12. (a) Comparison of measured and simulated motor currents for rear wheel and (b) corresponding states of RHyMo while its middle wheel climbs
the step. (For interpretation of the references to color in the text, the reader is referred to the web version of this article.)
34 D. Choi et al. / Mechanism and Machine Theory 114 (2017) 20–37

Fig. 13. Snapshots of RHyMo’s climbing the step with adjusting the contact angle according to its posture.

role in predicting the key characteristics of dynamic behavior of RHyMo climbing a step. Of course, it is very important to
derive a reliable model for the RHyMo for further researches.
Also, the region A is observed in Fig. 12(a), where the motor current for the rear wheel is drastically decreased from
2.6 A to 1.2 A after the middle wheel climbs the step. Fig. 12(b) describes the corresponding states of RHyMo while its
middle wheel climbs the step, where the angle ϕ between the links CI and DI becomes reduced while the state of RHyMo
is changed from the state 1 to the state 4. Because of this kinematic characteristics of RHyMo, its rear wheel can naturally
move toward the step with less efforts and as a result, the motor torque for the rear wheel seems to be relatively decreased
although the RHyMo still moves forward. Note that when the wheel is lifted up, the measured motor current is about 1.2 A,
which implies that in the region A, the rear wheel cannot contribute to the movement of RHyMo.

4.3. Experiment using a step with adjusting contact angle

Another experiment is carried out, where the contact angle of front track of RHyMo is appropriately adjusted according
to its posture. Fig. 13 shows the snapshot of RHyMo’s climbing the step with changing the contact angle of front track.
During the experiment, the moving speed of RHyMo is 10 m/min. Initially, the contact angle of front track of RHyMo is set
to be 45° to make the side of track contact with the corner of step and then, the RHyMo moves forward with decreasing
the contact angle by rotating the front track in the counter clockwise direction as shown in Fig. 13. In fact, while the contact
angle of front track is adjusted (from Fig. 13b to d), for the simplicity, the track does not move but only the contact angle
is changed so that torque current remains constant during the time duration t in Fig. 14.
Fig. 14 shows two motor currents for the rear wheel obtained by the experiments, where the orange and blue lines
denote the motor currents measured without and with adjusting the contact angle, respectively. By adjusting the contact
angle, the resulting motor current for the rear wheel is significantly reduced from 6.2 A to 4.5 A by more than 27% as shown
in Fig. 14, which implies that it is possible to reduce the motor torque required to climb the step by properly adjusting the
contact angle of front track. Recall that in both experiments, the region A is observed, where the motor currents are quite
decreased right after the front wheel and the middle wheel are on the step as confirmed in Figs. 12(a) and 14.
From the viewpoint of energy consumption, instead of using a motor of large capacity for driving the front track of a
fixed contact angle, it is recommended to use two motors of small capacity so that one is used for driving the front track
and the other is used for adjusting the contact angle of the front track to climb the step efficiently by virtue of actively
adjusting the contact angle. Also, when the contact angle of track becomes negative on the tread of the step, the contact
point between the track and the tread of the step moves forward along the moving direction and as a result, the traction
forces from the middle/rear wheels can be effectively transferred to the front track to help the RHyMo climb the step more
quickly.
D. Choi et al. / Mechanism and Machine Theory 114 (2017) 20–37 35

Fig. 14. Comparison of motor currents for rear wheel with/without adjusting the contact angle.

5. Conclusion

For the mobile robot which moves at a relatively high speed, it is very important to consider its dynamic effects such
as speed and acceleration in order to accurately obtain the forces and torques acting on the wheels and joints. In this
study, the position as well as the posture of links of proposed mobile robot RHyMo are first calculated via the kinematic
analysis. By using posture and position data, the linear/angular speeds and accelerations of links and wheels of RHyMo are
calculated and used for the force and moment equilibrium equations for the RHyMo. In this study, the kineto-static analysis
is applied to the RHyMo climbing the 150 mm high step with the moving speed of 10 m/min and the traction, normal forces
and moments required for the RHyMo to climb the step are calculated by using the concept of required friction coefficient,
which shows that the dynamic effects of RHyMo are well reflected by the kineto-static analysis compared to the quasi-static
analysis. Finally, in order to reduce the motor torque for rear wheel of RHyMo to climb the step, the contact angle of front
track of RHyMo is searched on the basis of kineto-static simulations. The strategy for adjusting the contact angle of front
track of RHyMo is investigated to improve the energy efficiency of the motor torque through the experiments using the
RHyMo. The experimental results show that the trend of torque for rear wheel simulated by the kineto-static analysis is
quite similar to that of real motor current and that the proposed adjustment of contact angle can significantly reduce the
required motor torque for the rear wheel. Presently, we keep working on deriving the exact model for the RHyMo to describe
its dynamic behavior with high fidelity and in the near future, based on this model, we will perform the optimization of
contact angle to significantly reduce the required torque irrespective of trajectory of terrain.

Acknowledgment

This work was supported by Basic Science Research Program through the National Research Foundation of Korea (NRF)
grants funded by the Ministry of Education (NRF 2014R1A1A2055668 and NRF 2016R1D1A1B03935516).

Appendix

For the simplicity of expression, let sαi and cαi denote sin α i and cos α i , respectively. By summing Eqs. (6), (9), (12), (15),
(18), (21) and (24), the following relation is obtained.

7
−sα1 · N1 + (μ2 · cα2 − sα2 )N2 + (μ3 · cα3 − sα3 )N3 + cα1 · T1 = mi aix + Mb a5x (29)
i=1

Similarly, by summing Eqs. (7), (10), (13), (16), (19), (22) and (25), the following equation is derived:

7
cα1 · N1 + (μ2 · sα2 + cα2 )N2 + (μ3 · sα3 + cα3 )N3 + sα1 · T1 = mi a iy + Mb a5y (30)
i=1

where ai = aiy + g. Substituting Eqs. (18) and (19) into Eq. (20) gives
y

F F
FDI x FIx − FDIyy FIy = I5 ω˙ 5 + IMb ω˙ 5 + (m5 a5x + Mb a5x )5Dx + (m5 a5y + Mb a5y )5Dy + Mb gM
F F bg
5
(31)
x
36 D. Choi et al. / Mechanism and Machine Theory 114 (2017) 20–37

F F F FDy FD FI
where FDx = 5Dx + 5Ix and F = 5 y + 5 y . By combining Eqs. (12), (13), (21), (22), (24) and (25), the following rela-
Ix Iy
tions are derived.

FHx = (μ3 · cα3 − sα3 )N3 − m3 a3x − FIx − m7 a7x


(32)
FHy = (μ3 · sα3 + cα3 )N3 − m3 a 3y − FIy − m7 a 7y

FFx = FHx − m6 a6x


(33)
FFy = FHy − m6 a 6y

By substituting Eqs. (32) and (33) into Eq. (23), the following equation is obtained:
 FH FH FH
F F F
sα3 · FHF x − cα3 · FF y − μ3 (cα3 · FHF x + sα3 · FF y ) N3 + FHF x FIx + FF y FIy
x y
FF
x y x y
FH
(34)
= I6 ω˙ 6 − m6 a6x · 6Fx − m6 a 6y · 6 y − (m3 a3x + m7 a7x )FHF x − (m3 a 3y + m7 a 7y )FF y
F F
x y

F F F FHy FH FF
where FHx = 6Hx + 6Fx and F = 6 y + 6 y . By substituting Eqs. (12), (13), (14) and (32) into Eq. (26), the following
Fx Fy
equation is obtained:
 F F F
μ3 (cα3 · FFH3xx − sα3 · FH3yy + r3 ) − sα3 · FFH3xx − cα3 · FH3yy N3 + FFIHxx FIx + FIHyy FIy
FH FH
(35)
= I3 ω˙ 3 + I7 ω˙ 7 + m3 a3x · FH3 x − m3 a 3y · F3 y + m7 a7x · 7Hx − m7 a 7y · 7 y
F F
x y

F F F FHy FH F3
where FHx = 7Hx + 73x and F = 7 y + 7 y . Substituting Eqs. (6)–(11), (18), (19) and (33) into Eq. (17) produces the
3x 3y
following equation.
F
 F
F F
(−sα1 · 41x + cα1 · 41y )N1 + μ2 r2 + (μ2 · cα2 − sα2 )42x − (μ2 · sα2 + cα2 )42y N2
 FF F1
F F
+ −(μ3 · sα3 + cα3 )4Fx − (μ3 · cα3 − sα3 )4 y N3 + (r1 + cα1 · 41x + sα1 · 4 y )T1
FF FD F1 (36)
+ m1 a1x · 41x + m1 a 1y · 4 y
F F F
+(4Fx − 4Dx )FIx + (4 y + 4 y )FIy = I1 ω˙ 1 + I2 ω˙ 2 + I4 ω˙ 4 + m1 g · m
1
1g

F2y FDy
− m2 a 2y ·  + (m5 a 5y + Mb a 5y )
F2x FDx
+m2 a2x ·  4 4
− (m5 a5x + Mb a5x ) 4 4
FF
− (m3 a 3y + m6 a 6y + m7 a 7y )4 y
FFx
−(m3 a3x + m6 a6x + m7 a7x ) 4

Let aij and bij denote the entries in the ith row and jth column of the matrices A6 × 6 and B6 × 1 , respectively. Then, from
Eqs. (29)–(31), (34)–(36), aij and bij are given by
a11 = −sα1 , a12 = −sα2 + μ2 · cα2 , a13 = −sα3 + μ3 · cα3 , a14 = cα1 , a15 = a16 = 0,
a21 = cα1 , a22 = cα2 + μ2 · sα2 , a23 = cα3 + μ3 · sα3 , a24 = sα1 , a25 = a26 = 0,
F FD F FH
a31 = a32 = a33 = a34 = 0, a35 = FDx , a35 = −F y , a41 = a42 = a44 = 0, a45 = FHx , a46 = F y ,
Ix Iy Fx Fy
F FHy F FH F FI
a43 = sα3 · FHx − cα3 · F − μ3 (cα3 · FHx + sα3 · F y ), a51 = a52 = a54 = 0, a55 = FIx , a56 = F y ,
Fx Fy Fx Fy Hx Hy
F FHy F FFy F F FFy FDy
a53 = μ3 (cα3 · FH3 x − sα 3 ·  F
3y
+ r3 ) − sα3 · FHx
3x
− cα3 · F , a65 =
3y
4Fx − 4Dx , a66 = 4 + 4 ,
x
F F1y F F2 F F2
a61 = −sα1 · 41x+ cα1 · 4 , a62 = μ2 (cα2 · 42x − sα2 · 4 y + r2 ) − sα2 · 42x − cα2 · 4 y ,
F FF F FF F F1
a63 = μ3 (−cα3 · 4Fx − sα3 · 4 y ) + sα3 · 4Fx − cα3 · 4 y , a64 = r1 + cα1 · 41x + sα1 · 4 y .
7 7
b11 = mi aix + Mb a5x ,b21 = mi a iy + Mb a5y ,
i=1 i=1
F FD
b31 = (I5 + IM )ω˙ 5 + (m5 a5x + Mb a5x )5Dx − (m5 a5y + Mb a5y )5 y ,
F FHy F FF
b41 = I6 ω˙ 6 − (m3 a3x + m7 a7x )FHx − (m3 a3y + m7 a7y )F − m6 a6x 6Fx − m6 a6y 6 y ,
Fx Fy
F FHy F FH
b51 = I3 ω˙ 3 + I7 ω˙ 7 + m3 a3x FHx − m3 a3y F + m7 a7x 7Hx − m7 a7y 7 y ,
3x 3y
F F1 F F2 F
b61 = I1 ω˙ 1 + I2 ω˙ 2 + I4 ω˙ 4 + m1 g1 1 + m1 a1x 41x + m1 a 1y 4 y + m2 a2x 42x − m2 a 2y 4 y − (m3 a3x + m6 a6x + m7 a7x )4Fx
m g

FFy F FDy
−(m3 a 3y + m6 a 6y + m7 a 7y )4 − (m5 a5x + Mb a5x )4Dx + (m5 a 5y + Mb a 5y )4 .

Supplementary materials

Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.mechmachtheory.
2017.03.018.
D. Choi et al. / Mechanism and Machine Theory 114 (2017) 20–37 37

References

[1] T. Thueer, R. Siegwart, Mobility evaluation of wheeled all-terrain robots, Robot. Autonom. Syst. 58 (2010) 508–519.
[2] B. Ghotbi, F. Gonzalez, J. Kovecses, J. Angeles, A novel concept for analysis and performance evaluation of wheeled rovers, Mech. Mach. Theory 83
(2015) 137–151.
[3] D. Kim, H. Hong, H.S. Kim, J. Kim, Optimal design and kinetic analysis of a stair-climbing mobile robot with Rocker–Bogie mechanism, Mech. Mach.
Theory 50 (2012) 90–108.
[4] R. Vidoni, M. Bietresato, A. Gasparetto, F. Mazzetto, Evaluation and stability comparison of different vehicle configurations for robotic agricultural
operations on side-slopes, Biosyst. Eng. 129 (2015) 197–211.
[5] P. Ben-Tzvi, A.A. Goldenberg, J.W. Zu, Design, simulations and optimization of a tracked mobile robot manipulator with hybrid locomotion and manip-
ulation capabilities, IEEE International Conference on Robotics and Automation, 2008.
[6] S. Jung, D. Choi, H.S. Kim, J. Kim, Trajectory generation algorithm for smooth movement of a hybrid-type robot Rocker–Pillar, J. Mech. Sci. Technol. 30
(2016) 5217–5224.
[7] K. Iagnemma, S. Dubowsky, Mobile Robots in Rough Terrain, Springer, Germany, 2004.
[8] R. Siegwart, P. Lamon, T. Estier, M. Lauria, R. Piguet, An innovative design for wheeled locomotion in rough terrain, Robot. Autonom. Syst. 40 (2002)
151–162.
[9] T. Thueer, P. Lamon, A. Krebs, R. Siegwart, Crab – exploration rover with advanced obstacle negotiation capabilities, The 9th ESA Workshop on Ad-
vanced Space Technologies for Robotics and Automation, 2006.
[10] C.K. Woo, D.H. Choi, S. Yoon, S.H. Kim, Y.K. Kwak, Optimal design of a new wheeled mobile robot based on a kinetic analysis of the stair climbing
states, J. Intell. Robotic Syst. 49 (2007) 325–354.
[11] D. Choi, J.R. Kim, S. Cho, S. Jung, J. Kim, Rocker-Pillar: design of the rough terrain mobile robot platform with caterpillar tracks and rocker bogie
mechanism, IEEE/RSJ International Conference on Intelligent Robots and Systems, 2012.
[12] H. Hong, D. Kim, H.S. Kim, S. Lee, J. Kim, Contact angle estimation and composite locomotive strategy of a stair-climbing mobile platform, Robot.
Comput.-Integrat. Manuf. 29 (2013) 367–381.
[13] V. Kucherenko, A. Bogatchev, M. Van Winnendael, Chassis concepts for the ExoMars rover, The 8th ESA Workshop on Advanced Space Technologies
for Robotics and Automation, 2004.
[14] R.A. Lindemann, C.J. Voorhees, Mars exploration rover mobility assembly design, test and performance, IEEE International Conference on Systems, Man
and Cybernetics, 2005.
[15] D. DeVon, T. Bretl, Kinematic and dynamic control of a wheeled mobile robot, IEEE/RSJ International Conference on Intelligent Robots and Systems,
2007.
[16] R. Lenain, B. Thuilot, O. Hach, P. Martinet, High-speed mobile robot control in off-road conditions: a multi-model based adaptive approach, IEEE
International Conference on Robotics and Automation, 2011.
[17] S.K. Jun, G.D. White, V.N. Krobi, Kinetostatic design considerations for an articulated leg-wheel locomotion subsystem, J. Dyn. Syst. Measur. Control
128 (2006) 112–121.
[18] A. Alamdari, V.N. Krovi, Kinetostatic optimization for an adjustable four-bar based articulated leg-wheel subsystem, IEEE/RSJ International Conference
on Intelligent Robots and Systems, 2014.
[19] A. Alamdari, V.N. Krovi, Design of articulated leg-wheel subsystem by kinetostatic optimization, Mech. Mach. Theory 10 (2016) 222–234.
[20] Z. Shiller, M.P. Mann, Dynamic stability of off-road vehicles, IEEE/RSJ International Conference on Intelligent Robots and Systems, 2004.
[21] Z. Shiller, M.P. Mann, D. Rubinstein, Dynamic stability of off-road vehicles considering a longitudinal terramechanics model, IEEE International Confer-
ence on Robotics and Automation, 2007.
[22] G.H. Golub, C.F. Van Loan, Matrix Computations, John Hopkins Press, 1989.
[23] D. Choi, Y. Kim, S. Jung, J. Kim, H.S. Kim, A new mobile platform (RHyMo) for smooth movement on rugged terrain, IEEE/ASME Trans. Mechatron. 21
(2016) 1303–1314.
[24] M. Minami, T. Asakura, N. Fujiwara, K. Kanbara, Inverse dynamics compensation method for PWS mobile manipulators, JSME Int. J. Series C, Mech.
Syst. Mach. Elements Manuf. 40 (1997) 291–298.
[25] E.M. Schearer, Y. Liao, E.J. Perreault, M.C. Tresch, W.D. Memberg, R.F. Kirsch, K.M. Lynch, Identifying inverse human arm dynamics using a robotic
testbed, IEEE/RSJ International Conference on Intelligent Robots and Systems, 2014.
[26] Y. Li, S. Staicu, Inverse dynamics of a 3-PRC parallel kinematic machine, Nonlinear Dyn. 67 (2012) 1031–1041.
[27] http://www.skf.com/group/products/bearings-units-housings/ball-bearings/principles/friction/estimating-frictional-moment/.
[28] G.F. Franklin, J.D. Powell, A. Emami-Naeini, Feedback Control of Dynamic Systems, 6th ed., Prentice Hall, 2009.
[29] M. Ruderman, Tracking control of motor drives using feedforward friction observer, IEEE Trans. Ind. Electron. 61 (2014) 3727–3735.

You might also like