You are on page 1of 9

Mario E.

Herrera-Cordero
Posgrade Division,
Universidad Tecnologica de la Mixteca,
Oaxaca 69000, Mexico
Cosimulation and Control of a
e-mail: herrera.enrique.09@gs.utm.mx
Single-Wheel Pendulum Mobile
Manuel Arias-Montiel1
Institute of Electronics and Mechatronics,
Universidad Tecnologica de la Mixteca,
Robot
Oaxaca 69000, Mexico Cosimulation is widely used as a powerful tool for performance evaluation of systems

Downloaded from http://asmedigitalcollection.asme.org/mechanismsrobotics/article-pdf/13/5/050909/6726813/jmr_13_5_050909.pdf by Concordia University user on 22 September 2021


e-mail: mam@mixteco.utm.mx design. This approach presents advantages over traditional design methodologies for
saving money and time in the development process and the possibility of evaluating
Marco Ceccarelli rapidly design alternatives by using virtual prototypes. This article presents an ADAMS/
LARM2: Laboratory of Robot Mechatronics, MATLAB cosimulation for the dynamics and control of a Single-Wheel pendulum ROBot
University of Rome-Tor Vergata, (SWROB) with inertial locomotion actuation to characterize design solutions by means of
00133 Rome, Italy validation of analytical results. The obtained results by the proposed cosimulation show
e-mail: marco.ceccarelli@uniroma2.it a significant performance based on the analytical and programming efforts in characteriz-
ing and simulating the designed system model. Moreover, open-loop experimental results
Esther Lugo-Gonzalez are presented to validate both the analytical model and the virtual prototype.
Institute of Electronics and Mechatronics, [DOI: 10.1115/1.4051359]
Universidad Tecnologica de la Mixteca,
Oaxaca 69000, Mexico Keywords: control, dynamics, mobile robots, robot design
e-mail: elugog@mixteco.utm.mx

1 Introduction application of specialized software packages that allow the use of


more sophisticated models or virtual prototypes as mentioned in
Currently, mobile robots address still a great interest for new and
Ref. [23]. Cosimulation allows designers to examine different kine-
enhanced solutions regarding different topics: locomotion, percep-
matic and dynamical aspects of a system and enables an automatic
tion, control system, and navigation [1–3]. To reduce the size and
evaluation of a much larger design alternatives when compared to a
the required space for operating mobile robots, different solutions
manual trial-and-error approach [24]. The cosimulation approach
for single-wheel robots have been proposed, i.e., single spherical
reduces the time that is required to design a complex system and
wheel mobile robots or ballbots [4–6], spherical robots [7–9], and
saves money by giving ability to modify the system for the user
single-wheel or unicycle robots [10–17]. Single-wheel robots
requirements before it is fabricated as mentioned in Ref. [25].
have compact sizes, and they are more flexible than other
Design of robotic devices involves mechanical, electrical, and elec-
wheeled robots because they can achieve zero rotation radius
tronic components as well as hardware and software modules, and
since there is only one point of contact with the surface. Because
a modeling approach allows a combination of separate domain-
of these characteristics, single-wheel robots have potential applica-
specific elements to create a complete model of the design by
tions such as terrain exploration and load transportation in narrow
collaborative exchange of information including the simulation
spaces as pointed out in Ref. [11]. However, a single-wheel
parameters, control signals, and system events [24]. Cosimulation
mobile robot is quite a challenging dynamical system from a
has been applied in the design and analysis of robotic manipulators
control point of view because it can easily fall down [16]. Different
like in Refs. [23,25–27] and mobile robots like in Refs. [24,28–30],
approaches have been proposed to solve the problem of lateral equi-
and the use of these computational tools has proved the advantages
librium in this kind of robot: horizontal inertial wheels [12], lateral
mentioned earlier as pointed out in Ref. [27].
inertial wheels [13], and high-speed gyroscopes [14]. Control stra-
This article is an extension of the work reported in Ref. [31]. The
tegies for single-wheel and unicycle robots are also developed for
novel contributions in this study are focused on the use of properly
straight trajectories as proposed by Lee et al. [15] and Vasudevan
elaborated simulation codes for multibody systems in ADAMS VIEW
et al. [18]. In single-wheel robot design and control, a representative
for a dynamic analysis of a Single-Wheel ROBot (SWROB) with
investigation is the work by Jung et al. In Ref. [16], a mechatronics
inertial locomotion actuation that is based on the controlled
approach is proposed for the development of the GYROBO regard-
motion of a simple pendulum. Moreover, a cosimulation ADAMS/
ing the system integration. It allows implementation of linear con-
MATLAB SIMULINK is used to evaluate the performance of a designed
trollers for both roll and yaw angles. In Ref. [17], the stability of
control algorithm that is based on the root locus method to regulate
the vertical mode of GYROBO is analyzed, and its balancing
the linear velocity of the robot. The obtained results are compared
control performance is tested to verify the stability bandwidth.
with the preliminary results that are partially reported in
Recently, Forouhar et al. [19] presented a novel single-wheel
Ref. [31], showing a considerable reduction efforts for analytical
robot as based with flywheels motors, pendulum, and wheel. The
modeling and numerical computation in designing and simulating
dynamics of the robot was formulated using the Euler–Lagrange
the proposed system model with proper performance. Finally, an
equations to propose a control system as based on fuzzy logic to sta-
experimental prototype of the robot was built, and tests of the open-
bilize the robot from falling only.
loop control system are presented to validate the numerical results
In the last years, cosimulation has become a powerful tool for the
that are obtained in the virtual prototype as also a characterization
development of mechatronic systems [20–22]. It consists in
of the proposed design of a single-wheel mobile robot.

1
Corresponding author.
Contributed by the Mechanisms and Robotics Committee of ASME for publication 2 Materials and Methods
in the JOURNAL OF MECHANISMS AND ROBOTICS. Manuscript received March 7, 2021; final
manuscript received May 20, 2021; published online July 6, 2021. Assoc. Editor: 2.1 Robot Design and Virtual Prototype. The robot design is
Renato Vidoni. based on the barycenter offset driving principle. In the proposed

Journal of Mechanisms and Robotics Copyright © 2021 by ASME OCTOBER 2021, Vol. 13 / 050909-1
SWROB system, the mass distribution of the robots is shifted by an
internal mechanism to move the robot from one equilibrium posi-
tion to another. This principle has been implemented in several
locomotion systems of spherical robots as pointed out in Ref. [7].
In this article, the robot is only designed for forward and back-
ward movement on a straight trajectory, and the mass distribution
is shifted by a movement of an inertial pendulum driven by a CD
micromotor with internal spur gears mechanisms. The mechanical
elements of the SWROB are shown in Fig. 1. The robot structure
is mainly composed by robot housing (1) considered as a cylindrical
tube, and in practice, it was obtained by eight 3D-printed pieces;
two base discs (2), which is used as a support for all the mechanical

Downloaded from http://asmedigitalcollection.asme.org/mechanismsrobotics/article-pdf/13/5/050909/6726813/jmr_13_5_050909.pdf by Concordia University user on 22 September 2021


and electronic components for the inertial locomotion mechanism; Fig. 2 A SWROB model of Fig. 1 in ADAMS VIEW environment
pendulum (4) formed by both a sphere and a stem made of steel,
and six bolts with hollow cylinders symmetrically separated
around the axis (6) used to contain the batteries to provide energy Table 1 Main characteristics of the SWROB in Figs. 1 and 2
to the motor; the contact between base discs and base pendulum
(3) has little friction, and no slip between wheel and the ground Element Mass (g) Moment of inertia (g · cm2) Radius (cm)
(5) is considered. The main characteristics of the proposed design
can be summarized as follows: Housing 583.44 73884.32 13.50
Base discs 70.42 1548.73 −
• Traction surfaces on both sides of the wheel by placing two Bolts 1.87 0.35 1.64
pinions on the extended shaft of the motor. This action Pendulum 122.36 586.61 10.34
cancels the effect of the torques that can be generated in the Pinion Negligible Negligible 0.75
normal direction to the motion plane. Gear 4.20 7112.64 3.97
• The weight of the mechanical elements is distributed in a sym-
metrical way with respect to rotation axis of the robot.
• Holding slides are added on both sides of the wheel to ensure
the stability of locomotion. (3) Eight bolts symmetrically distributed with respect to the rota-
tion axis.
Preliminary design characteristics are presented in Ref. [32] with (4) Friction between the two base discs and pendulum is
the development and validation by SOLIDWORKS simulation of the neglected.
system dynamic model. However, that model is obtained based (5) The mass and moment of inertia of the pinion are neglected
on a conceptual design by omitting components that are included since those values are considerably smaller than those values
in the detailed here-proposed design and in the new experimental of the other mechanical elements.
prototype.
The virtual prototype of the SWROB design was developed with Figure 3 shows an example of computed results for the
all the components in ADAMS VIEW software [33], as shown in Fig. 2 angular velocity of the SWROB with a constant input torque of
considering values for the physical parameters in Table 1 as well as 2N · mm, where the angular velocity slowly reach the steady state
the geometrical relationships and the mechanical constrains as in the in about 17 s.
following:
(1) No slipping between the wheel and the ground. 2.2 Analytical System Dynamics. The dynamic model of the
(2) The pendulum is composed of a sphere and a solid rod as one SWROB is obtained as based on mechanical elements in Fig. 1 and
solid body. Table 1 with the mentioned in Sec. 2.1.
Figure 4 shows a free-body diagram of a simplified view of the
system, where θ is the robot angular displacement, τ is the input
torque provided by the micromotor, θ2 = θ − Rα is the pendulum
angle, α is the pinion angular displacement, and R is the transmis-
sion ratio of the internal spur gears mechanism.
The linear velocity of the robot is given by

Ẋ = Rc θ̇ (1)
where Rc is the radius of the robot housing and θ̇ is the robot angular
velocity.
The system’s kinetic and potential energies can be computed by
an analysis with elements depending on the variable θ and sepa-
rately, with the elements depending on θ and α as follows. The ele-
ments depending on the variable θ are the robot housing, the base
discs, the bolts, and the internal gears (see Fig. 5). Their kinetic
and potential energies can be expressed as follows:
1
T1 = [mc R2c + md R2c + mcor R2c + Ic + Icor
2
2
+ 8(mbolt (L2q + R2c ) + Ibolt )]θ̇ (2)

2
V1 = mc gRc + md g(Rc − Rd ) + mcor g(Rc − Rcor ) + 8mbolt gLq θ̇ (3)
Fig. 1 A conceptual design of the SWROB: Robot housing (1),
two base discs (2), base pendulum (3), pendulum (4), ground- where Lq is the distance from the bolt axis to the rotation axis of the
contact disc (5), and eight bolts (6) wheel, mc is the robot housing mass, md is the base discs mass, mcor

050909-2 / Vol. 13, OCTOBER 2021 Transactions of the ASME


Downloaded from http://asmedigitalcollection.asme.org/mechanismsrobotics/article-pdf/13/5/050909/6726813/jmr_13_5_050909.pdf by Concordia University user on 22 September 2021
Fig. 3 Computed results of ADAMS simulation of SWROB in terms of angular velocity
with a constant input torque

is the internal gears mass, mbolt is a bolt mass, and Ic, Id, Icor, and
Ibolt denote the corresponding moment of inertia.
The elements depending on the variables α and θ are the pendu-
lum and its base (the composed pendulum) with mass center in Ctm
with the coordinates Xctm and Yctm as shown in Fig. 5. The kinetic
and potential energies of the pendulum can be expressed as follows:
1
T2 = mcomp R2c θ̇ + mcomp ẊDctm cos(θ2 )θ˙2
2
2
1 1
+ mcomp D2ctm θ˙2 + Ictm θ˙2
2 2
(4)
2 2

V2 = mcomp gDctm (1 − cos(θ2 )) (5)


where mcomp is the mass of the composed pendulum and Ictm is the
moment of inertia of the pendulum.
The total kinetic and potential energies of the SWROB can be
obtained by Eqs. (2)–(5), in the form
1
Ttotal = [mc R2c + md R2c + mcor R2c + Ic + Icor
2
2
+ 8(mbolt (L2q + R2c ) + Ibolt ) + mcomp R2c ]θ̇
Fig. 4 A free-body diagram of the SWROB 1 1
+ mcomp ẊDctm cos(θ2 )θ˙2 + mcomp D2ctm θ˙2 + Ictm θ˙2
2 2
(6)
2 2

Vtotal = mc gRc + md g(Rc − Rd ) + mcor g(Rc − Rcor )


2
+ 8mbolt gLq θ̇ + mcomp gDctm (1 − cos (θ2 )) (7)
The Euler–Lagrange formulation can be applied to obtain a
dynamic model of the SWROB as follows:

θ̈ = {mcomp Dctm Rc sin(bα − θ)(bα̇ + θ̇)2


+ mcomp gDcomp sin(bα − θ) − [mcomp Dctm Rc cos(bα − θ)
1
− Icomp − mcomp D2comp ]bα̈ − μθ̇} (8)
W

α̈ = − [mcomp Dctm Rc cos(bα − θ) − Ip − mcomp D2ctm ]θ̈
b
 1
−mcomp gDctm sin(bα − θ) (9)
[Icomp + mcomp D2ctm ]b
where τ is the input torque that is provided by the motor, μ is the
R pinion
friction coefficient between the wheel and the ground, b = ,
Rcor
Fig. 5 Variables for the calculation of kinetic and potential ener- W = e − 2mcomp Dctm Rc cos(bα − θ) + Icomp + mcomp Dctm , and e =
2

gies of the pendulum mc R2c + md R2c + mcor R2c + 4(mbolt (L2q + R2c )) + Ic + Id + Icor + 8Ibolt .

Journal of Mechanisms and Robotics OCTOBER 2021, Vol. 13 / 050909-3


Equations (8) and (9) can be used to evaluate the system dynamic Table 2 Numerical parameters for the linearized model and
performance with different values of the input torque, to verify sta- velocity transfer function
bility and design a control law to modify the behavior of the system
output, as well as to quantify the needed torque provided by the Parameter Value Parameter Value
motors to drive the robot.
a11 0 a31 0
a12 1 a32 0
a13 0 a33 0
2.3 Linearization, Transfer Function, and Velocity a14 0 a34 1
a21 −3.795 a41 486.090
Control. From Eqs. (8) and (9), it is possible to linearize the a22 −0.303 a42 0.400
system model around an operating point. This point is considered a23 0.716 a43 −91.715
with the pendulum in a vertical position down (which is also an a24 0 a44 0
equilibrium position), with θ2 = 0 and θ = bα. Moreover, the equi-

Downloaded from http://asmedigitalcollection.asme.org/mechanismsrobotics/article-pdf/13/5/050909/6726813/jmr_13_5_050909.pdf by Concordia University user on 22 September 2021


b0 11770 a0 27.541
librium torque input is 0, and as a consequence, the system has a b1 0 a1 95.510
constant velocity, θ̈ = 0 and α̈ = 0. The state variables can be b2 −32.307 a2 0.303
defined as follows: x1 = θeq, x2 = θ̇eq , x3 = αeq, x4 = α̇eq , and Ueq = 0.
By using the equilibrium conditions, it is possible to linearize the
model by using the Taylor series expansion [34]. The linearized
model in state space can be expressed as follows: It can be noted that the second row from the right-hand side of
Eq. (13) corresponds to the transfer function of the SWROB velo-
ẋ = Ax + Bu (10) city as follows:

y = Cx + Du b2 s2 + b1 s + b0
(11) G2 (s) = (14)
s3 + a2 s2 + a1 s + a0
where
Numerical values for the parameters of the linearized system model
⎡ ⎤ ⎡ ⎤ in Eqs. (10) and (11) and for the robot velocity transfer function in
x1δ a11 a12 a13 a14
⎢x ⎥ ⎢a a24 ⎥
Eq. (14) are presented in Table 2, and they are obtained by consid-
⎢ 2δ ⎥ ⎢ 21 a22 a23 ⎥
x = ⎢ ⎥, A=⎢ ⎥ ering the numerical values of the robot characteristics from Table 1.
⎣ x3δ ⎦ ⎣ a31 a32 a33 a34 ⎦ The transfer function in Eq. (14) can be used to design a control
x4δ a41 a42 a43 a44 scheme for the single-wheel robot velocity. To validate this function,
⎡ ⎤ ⎡ ⎤ ⎡ ⎤ a numerical simulation with a constant input torque of 2N · mm is
b11 1 0 0 0 0 carried out. The computed results are compared with the nonlinear
⎢b ⎥ ⎢0 1 0 0⎥ ⎢0⎥ system response, and this comparison is presented in Fig. 6.
⎢ 21 ⎥ ⎢ ⎥ ⎢ ⎥
B=⎢ ⎥, u = τ, C=⎢ ⎥, D = ⎢ ⎥ As one can appreciate, the obtained response from the transfer
⎣ b31 ⎦ ⎣0 0 1 0⎦ ⎣0⎦
function is practically the same as one the obtained from nonlinear
b41 0 0 0 1 0 equations. These results validate the linearization process and
From the linearized system with Eqs. (10) and (11), the transfer justify the use of the proposed transfer function to synthesize the
function can be obtained as follows: velocity control by linear techniques.
Before designing a controller, it is convenient to analyze the pole
G(s) = C(sI − A)−1 B + D (12) and zeros in open and closed loop. Figure 7 shows the computed
distribution of pole and zeros in open loop. The corresponding
where open-loop system is stable due to both poles having negative real
⎡ ⎤ parts.
G1 (s) The closed-loop transfer function can be expressed as follows:
⎢ ⎥
⎢ ⎥
⎢ G2 (s) ⎥ C(s) G2 (s) −32.31s2 + 11772
⎢ ⎥ = = 3 (15)
G(s) = ⎢


⎥ (13) R(s) 1 + G2 (s)H(s) s − 32s2 + 95.51s + 11800
⎢ G3 (s) ⎥
⎢ ⎥ From the transfer function in Eq. (15), it is possible to obtain the
⎣ ⎦
location of pole and zeros in a closed loop as shown in Fig. 8. There
G4 (s) is a pair of conjugated poles in the right semi-plane, which means

Fig. 6 A comparison between linearized model and nonlinear formulation

050909-4 / Vol. 13, OCTOBER 2021 Transactions of the ASME


Downloaded from http://asmedigitalcollection.asme.org/mechanismsrobotics/article-pdf/13/5/050909/6726813/jmr_13_5_050909.pdf by Concordia University user on 22 September 2021
Fig. 9 Computed pole and zeros for the compensated
Fig. 7 Computed pole and zeros placement for the open-loop closed-loop control
system

rad
frequency have been assumed as ζ = 0.5 and ωn = 1 , respec-
s
tively. With these values,

the dominant poles are identified by


s1,2 = ζωn ± 1 − ζ i ⇒ s1,2 = −0.5 ± 0.866i.
2

For this design case, the closed-loop instability is caused by the


dominant pole in the right semi-plane and the zero in the left semi-
plane (see Fig. 8). A direct compensator can be proposed to cancel
this pair of conjugated poles and the zero in the left side of the
complex plane by using
−32.307(s − 19.088)
G(s)s = G2 (s) ∗ G(s)supr = (16)
(s + 0.288)(s)
The proposed controller must compensate for the deficiency
angle that is introduced by a replacement of the dominant
closed-loop pole. By working out a graphic method as in
Ref. [35], the transfer function of the compensated closed-loop con-
Fig. 8 Computed pole and zeros for the proposed closed-loop trolled system can be given by
control system
−32.307(s − 19.088) K(s + 0.604)
GT = G(s)s ∗ G(s)comp = ∗ (17)
(s + 0.288)(s) s + 1.655
the closed-loop system is unstable even if there are variations in the
gain. where the calculated gain is K = 2.329 × 10−3.
It is possible to apply the root locus control technique to impose a The location of pole and zeros of the compensated closed-loop
desired behavior in a closed-loop controlled system. In this case, system is presented in Fig. 9, where the location of the dominant
the proposed values for the damping coefficient and natural pole in the desired location can be observed.

Fig. 10 A comparison of the computed results for linear velocity of the single-wheel
pendulum robot with a constant input torque and with the proposed controller

Journal of Mechanisms and Robotics OCTOBER 2021, Vol. 13 / 050909-5


Downloaded from http://asmedigitalcollection.asme.org/mechanismsrobotics/article-pdf/13/5/050909/6726813/jmr_13_5_050909.pdf by Concordia University user on 22 September 2021
Fig. 13 A comparison of computed results from cosimulation
and transfer function responses

Fig. 11 Simulation results for the required torque in the


where a Kalman filter is included to eliminate the high-frequency
designed SWROB control
noise in the output signal that can be generated by the solver algo-
rithm of ADAMS VIEW.
Figure 13 shows a comparison of the closed-loop controlled
The compensated closed-loop controlled system given by
system response as designed in the cosimulation procedure shown
Eq. (17) was numerically simulated to appreciate the robot velocity
in Fig. 12 and the closed-loop transfer function given in Eq. (17).
behavior. The obtained results from this simulation are presented in
The filtered response of cosimulation result follows the response
Fig. 10 with a comparison between the open-loop system with a
given by the transfer function in closed loop so that it gives an effi-
constant input torque and the controlled system, indicating the con-
cient operation of the designed virtual prototype as an alternative to
troller performance with an overshoot of about 26% and the settling
the dynamic model obtained by the analytical procedure.
time around 205% faster than the open-loop system. Moreover,
oscillations are suppressed in quite short time.
The compensated closed-loop controlled system response could 3.2 A Prototype and Experimental Results. A prototype was
be improved by modifying the requirements for the controller built to obtain experimental results to validate the virtual prototype
design. However, it is important to take into account the physical that has been designed with the cosimulation procedure. A PC was
limitations of the micromotor that is proposed to drive the used as control interface, and the communication between this and
SWROB. In Fig. 11, the required torque for the robot control is the on-board microcontroller (ATMega 328) was established by a
computed with a maximum required torque of 5.325N · mm to bluetooth antenna. The microcontroller receives data from a gyro-
select a specific micromotor with the adequate characteristics for scope that is used to measure the angular position of the robot
this application. and sends the information to the PC for the control algorithm in
Fig. 12. In addition, the microcontroller sends control signals that
are received from the PC to the micromotor driver to regulate the
input torque for the robot.
3 Results and Discussion The manufacture of most of the parts of SWROB was made by
3.1 Closed-Loop System From Cosimulation Results. By 3D printing, and the pendulum was constructed of steel to have
using the virtual prototype described in Sec. 2.1 an ADAMS/MATLAB the inertial properties that needed to drive the robot motion with
cosimulation is carried out as an alternative for an analysis of proper characteristics. Figure 14 shows the main components of
dynamics and velocity control of the SWROB. For this aim, the the robot’s mechanical system, and Fig. 15 illustrates the assembled
virtual prototype in ADAMS VIEW is exported to MATLAB SIMULINK final prototype.
environment where a block diagram with inputs and outputs is gen- Experiments of the open-loop system were carried out to
erated as shown in Fig. 12. In this case, the inputs are the torques compare the results with numerical simulation of the virtual proto-
that are provided by the micromotors and are generated by the type and to check the efficiency of the designed system. Figure 16
synthesized control algorithms in the previous section, and the shows the virtual prototype response and three different experimen-
output is the robot velocity, which is the variable to be controlled tal measurements, indicating the same value of velocity in steady

Fig. 12 A MATLAB block diagram for the cosimulation of SWROB in control design

050909-6 / Vol. 13, OCTOBER 2021 Transactions of the ASME


Downloaded from http://asmedigitalcollection.asme.org/mechanismsrobotics/article-pdf/13/5/050909/6726813/jmr_13_5_050909.pdf by Concordia University user on 22 September 2021
Fig. 14 Manufactured mechanical components for SWROB prototype: (a) base discs,
(b) pins and bolts, (c) pendulum-base assemble, and (d) housing

Fig. 16 Result of tests with real and virtual prototype in terms of


Fig. 15 The built prototype of the SWROB open-loop system response for a commanded velocity of 2.618
rad/s (150 deg/s)

state as well as the settling time. The experiments were carried out related to disturbances and noise in the gyroscope that is used to
with a constant input torque of 2N · mm and the SWROB running measure the angular position of the robot. Moreover, angular velo-
on a flat horizontal surface as shown in video snapshots of city of the SWROB is obtained by numerical differentiation of the
Fig. 17. However, oscillations with a large amplitude are observed angular position measured by the gyroscope located on the rotation
in the experimental measurements. These oscillations are mainly axis of the robot. This operation could introduce numerical noise to

Fig. 17 Video snapshots of an experiment with SWROB: (a) starting experiment,


(b) SWROB breaking inertia, (c) SWROB starting forward motion, (d) SWROB accelerat-
ing by pendulum inclination, (e) another perspective of SWROB motion, (f) SWROB
desaccelerating, (g) SWROB starting to slow down, (h) SWROB stopping, and (i) finish-
ing experiment

Journal of Mechanisms and Robotics OCTOBER 2021, Vol. 13 / 050909-7


Downloaded from http://asmedigitalcollection.asme.org/mechanismsrobotics/article-pdf/13/5/050909/6726813/jmr_13_5_050909.pdf by Concordia University user on 22 September 2021
Fig. 18 Result of tests with real and virtual prototype in terms of open-loop system
response for a commanded velocity of 1.55 rad/s (89 deg/s)

Fig. 19 Comparison between two tests of SWROB

the velocity signal. This oscillatory behavior can be observed also in implementation of the proposed controller can be planned to verify
Fig. 18, where experimental results for a lower angular velocity are the closed-loop performance.
presented.
In Fig. 19, a comparison between two experimental responses
with different commanded velocities is presented. In this figure, Conflict of Interest
one can observe that both experiments exhibit a similar dynamic There are no conflicts of interest.
behavior (settling time, oscillations about reference). These
similarities are derived from the control tuning, which is the same
for both tests.
Data Availability Statement
The datasets generated and supporting the findings of this article
are obtainable from the corresponding author upon reasonable
4 Conclusions request. The authors attest that all data for this study are included
This study presents extended results about control, cosimulation, in the paper.
and testing of a SWROB with inertial locomotion actuation. The use
of a virtual prototype in ADAMS VIEW software demonstrates accurate
results when compared with analytical models that are obtained by References
complex procedures. Moreover, the virtual prototype can be used [1] Rubio, F., Valero, F., and Llopis-Albert, C., 2019, “A Review of Mobile Robots:
for cosimulation procedure giving the possibility to prove different Concepts, Methods, Theoretical Framework, and Applications,” Int. J. Adv. Rob.
design alternatives by adding or removing elements or even by Syst., 16(2), pp. 1–22.
[2] Ceccarelli, M., and Kececi, E. F., 2015, Designs and Prototypes of Mobile Robots,
adjusting parameters values in the design model. Cosimulation ASME Press, New York.
approach can be also used to evaluate the controller performance [3] Kececi, E. F., and Ceccarelli, M., 2015, Mobile Robots for Dynamic
in a more realistic environment, which allows to visualize and Environments, ASME Press, New York.
modify the dynamic behavior of the controlled system. Finally, a [4] Nagarajan, U., Kantor, G., and Hollis, R., 2014, “The Ballbot: An
Omnidirectional Balancing Mobile Robot,” Int. J. Rob. Res., 33(6), pp. 917–930.
first prototype was constructed, and preliminary experimental [5] Navabi, H., Sadeghnejad, S., Ramezani, S., and Baltes, J., 2017, “Position Control
results of the open-loop system validate the virtual prototype that of the Single Spherical Wheel Mobile Robot by Using the Fuzzy Sliding Mode
is used for cosimulation procedure. As future work, an experimental Controller,” Adv. Fuzzy Syst., 2017, pp. 1–10.

050909-8 / Vol. 13, OCTOBER 2021 Transactions of the ASME


[6] Pham, D. B., Weon, I., and Lee, S., 2020, “Partial Feedback Linearization Back, M. ter Beek, A. Fantechi, and L. Semini, eds., 1st ed., Vol. 11865,
Double-Loop Control for a Pseudo-2d Ridable Ballbot,” Int. J. Control. Autom. Lecture Notes in Computer Science, Springer, Cham, Switzerland, pp. 40–55,
Syst., 18(5), pp. 1310–1323. Chapter 4.
[7] Chase, R., and Pandya, A., 2012, “A Review of Active Mechanical Driving [22] Gomes, C., Thule, C., Broman, D., Larsen, P. G., and Vangheluwe, H., 2018,
Principles of Spherical Robots,” Robotics, 1(1), pp. 3–23. “Co-Simulation: A Survey,” ACM Comput. Surveys, 51(3), pp. 1–33.
[8] DeJong, B. P., Karadogan, E., Yelamarthi, K., and Hasbany, J., 2017, “Design [23] Baran, D., and Lisowski, W., 2013, “Numerical Simulation and Co-Simulation in
and Analysis of a Four-Pendulum Omnidirectional Spherical Robot,” J. Intel. Analysis of Manipulator Dynamics,” Mech. Control, 32(4), pp. 129–135.
Rob. Syst., 86(1), pp. 3–15. [24] Christiansen, M. P., Larse, M. G., and Jorgensen, R. N., 2015, “Robotic Design
[9] Ivanova, T. B., Kilin, A. A., and Pivovarova, E. N., 2018, “Controlled Motion of a Choice Overview Using Co-Simulation and Design Space Exploration,”
Spherical Robot With Feedback,” J. Dyn. Control Syst., 24(3), pp. 497–510. Robotics, 4(4), pp. 398–420.
[10] Reina, G., Foglia, M. M., Millela, A., and Gentile, A., 2006, “Visual and [25] Sudharsan, J., and Karunamoorthy, L., 2016, “Path Planning and Co-Simulation
Tactile-Based Terrain Analysis Using a Cylindrical Mobile Bobot,” J. Dyn. Control of 8 Dof Anthropomorphic Robotic Arm,” Int. J. Simul. Model., 15(2),
Syst. Meas. Control., 128(1), pp. 165–170. pp. 302–312.
[11] Jin, H., Wang, T., Yu, F., Zhu, Y., Zhao, J., and Lee, J., 2016, “Unicycle Robot [26] Parthasarathy, T., Srinivasaragavan, V., and Santhanakrishnan, S., 2016,
Stabilized by the Effect of Gyroscopic Precession and Its Control Realization “Adams-Matlab Co-Simulation of a Serial Manipulator,” Proceedings of the

Downloaded from http://asmedigitalcollection.asme.org/mechanismsrobotics/article-pdf/13/5/050909/6726813/jmr_13_5_050909.pdf by Concordia University user on 22 September 2021


Based on Centrifugal Force Compensation,” IEEE/ASME Trans. Mech., 21(6), 3rd International Conference on Mechatronics and Mechanical Engineering
pp. 2737–2745. (ICMME 2016), Shanghai, China, Oct. 21–23, EDP Sciences, pp. 1–6.
[12] Vos, W. D., Flotow, A. H. V., and Zhuang, H., 1990, “Dynamics and Nonlinear [27] Sosa-Mendez, D., Lugo-Gonzalez, E., Arias-Montiel, M., and Garcia-Garcia,
Adaptive Control of an Autonomous Unicycle: Theory and Experiment,” R. A., 2017, “Adams-Matlab Co-Simulation for Kinematics, Dynamics, and
Proceedings of the 29th IEEE Conference on Decision and Control (CDC Control of the Stewart–Gough Platform,” Int. J. Adv. Rob. Syst., 14(4), pp. 1–10.
1990), Honolulu, HI, Dec. 5–7, IEEE, pp. 182–185. [28] Angel, L., Hernandez, C., and Diaz-Quintero, C., 2013, “Modeling, Simulation
[13] Jin, H., Hwang, J., and Lee, J., 2011, “A Balancing Control Strategy for a and Control of a Differential Steering Type Mobile Robot,” Proceedings of the
One-Wheel Pendulum Robot Based on Dynamic Model Decomposition: 32nd Chinese Control Conference, Xi’an, China, July 26–28, IEEE, pp. 8757–
Simulations and Experiments,” IEEE/ASME Trans. Mech., 16(4), pp. 763–768. 8762.
[14] Zhu, Y., Gao, Y., Xu, C., Zhao, J., Jin, H., and Lee, J., 2015, “Adaptive Control of [29] Ciszewsk, M., Buratowski, T., and Giergiel, M., 2018, “Modeling, Simulation
a Gyroscopically Stabilized Pendulum and Its Application to a Single-Wheel and Control of a Pipe Inspection Mobile Robot With an Active Adaptation
Pendulum Robot,” IEEE/ASME Trans. Mech., 20(5), pp. 2095–2106. System,” IFAC-PapersOnLine, 51(22), pp. 132–137.
[15] Lee, J., Han, S., and Lee, J., 2013, “Decoupled Dynamic Control for Pitch and [30] Han, J., Ren, G., and Li, D., 2017, “Co-Simulation of a Tracked Mobile Robot
Roll Axes of the Unicycle Robot,” IEEE. Trans. Ind. Electron., 60(9), Based on Recurdyn and Simulink,” Adv. Eng. Res., 130, pp. 1170–1174.
pp. 3814–3822. [31] Herrera-Cordero, M. E., Arias-Montiel, M., Ceccarelli, M., and Lugo-Gonzalez,
[16] Park, J. H., and Jung, S., 2013, “Development and Control of a Single- E., 2020, “On the Dynamics and Control of a Single-Wheel Robot With
Wheel Robot: Practical Mechatronics Approach,” Mechatronics, 23(6), Inertial Locomotion,” Industrial and Robotic Systems Proceedings of LASIRS
pp. 594–606. 2019, E. Hernandez, S. Keshtkar, and S. Valdez, eds., 1st ed., Vol. 86,
[17] Lee, S. D., and Jung, S., 2015, “Experimental Verification of Stability Region of Mechanisms and Machine Science, Springer, Cham, Switzerland, pp. 249–260,
Balancing a Single-Wheel Robot: An Inverted Stick Model Approach,” Chapter 24.
Proceedings of 41st Annual Conference of the IEEE Industrial Electronics [32] Herrera-Cordero, M. E., Arias-Montiel, M., and Lugo-Gonzalez, E., 2019,
Society (IECON 2015), Yokohama, Japan, Nov. 9–12, IEEE, pp. 4556–4561. “Design and Dynamic Modeling of a Novel Single-Wheel Pendulum Robot,”
[18] Vasudevan, H., Dollar, A. M., and Morrell, J. B., 2015, “Design for Control of Mechanism Design for Robotics. MEDER 2018, A. Gasparetto, and
Wheeled Inverted Pendulum Platforms,” ASME J. Mech. Rob., 7(4), pp. 1–12. M. Ceccarelli, eds., 1st ed., Vol. 66, Mechanisms and Machine Science,
[19] Forouhar, M., Abedin-Nasab, M. H., and Liu, G., 2020, “Introducing Gyrosym: A Springer, Cham, Switzerland, pp. 353–360, Chapter 42.
Single-Wheel Robot,” Int. J. Dyn. Control, 8(2), pp. 404–417. [33] McConville, J., 2015, Introduction to Mechanical System Simulation Using
[20] Fitzgerald, J., Pierce, K., and Larsen, P. G., 2014, “Co-Modelling and Adams, SDC Publications, New York.
Co-Simulation in the Engineering of Systems of Cyber-Physical Systems,” [34] Roubal, J., Husek, P., and Stecha, J., 2010, “Linearization: Students Forget the
Proceedings of 2014 9th International Conference on System of Systems Operating Point,” IEEE Trans. Edu., 53(3), pp. 413–418.
Engineering (SOSE), Glenelg, SA, Australia, June 9–13, IEEE, pp. 67–72. [35] Zanasi, R., and Cuoghi, S., 2011, “Direct Methods for the Synthesis of Pid
[21] Fitzgerald, J., Larsen, P. G., and Pierce, K., 2019, “Multi-Modelling and Compensators: Analytical and Graphical Design,” Proceedings of IECON 2011
Co-Simulation in the Engineering of Cyber-Physical Systems: Towards the —37th Annual Conference of the IEEE Industrial Electronics Society,
Digital Twin,” From Software Engineering to Formal Methods and Tools, and Melbourne, VIC, Australia, Nov. 7–10, IEEE, pp. 552–557.

Journal of Mechanisms and Robotics OCTOBER 2021, Vol. 13 / 050909-9

You might also like