SlideShare a Scribd company logo
1 Copyright © 2015 by ASME
DRAFT
Proceedings of the ASME 2015 International Mechanical Engineering Congress & Exposition
IMECE2015
November 13-19, 2015, Houston, Texas
IMECE2015-51551
A FOUR-PENDULUM OMNIDIRECTIONAL SPHERICAL ROBOT: DESIGN
ANALYSIS AND COMPARISON
Brian P. DeJong
Central Michigan University
Mount Pleasant, MI, USA
Kumar Yelamarthi
Central Michigan University
Mount Pleasant, MI, USA
Brad Bloxsom
Central Michigan University
Mount Pleasant, MI, USA
ABSTRACT
This paper presents a four-pendulum design for an
omnidirectional spherical robot. The robot moves by the
rotation of tetrahedrally-located pendulums to change the center
of mass, and can do so instantaneously in any direction. In that
respect, it is omnidirectional unlike the more common single-
pendulum designs. In the paper, the theoretical design is
discussed and contrasted to existing theoretical center-of-mass
designs in terms of directionality, torque arm, eccentricity, and
radius of gyration. The torque arm comparison is based on the
envelope of the dynamically-controlled mass (i.e., how far in
any direction can the center of mass be moved), while the
eccentricity and gyration criteria are based on homogeneity and
size of the mass moment of inertia (i.e., how much does the
design “wobble” while rolling; how much inertia does it have).
In these regards, the four-pendulum architecture is a strong
design – better than some in one respect, better than the others
in other respects. A prototype has been built and tested that
successfully implements the four-pendulum propulsion, and its
results are compared to theoretical ones.
INTRODUCTION
Spherical mobile robots are an interesting research area
with subtly complex dynamics and a variety of mechanisms [1-
3]. Various propulsion schemes have been used, and since one
scheme has not emerged clearly better than the rest, new
designs are still being presented. The ball-shaped robots are
intriguing because they have often have no externally-visible
sources of propulsion and yet can navigate the non-holonomic
rolling-ball-on-plane space. They have application in patrol
and surveillance [4], extraterrestrial exploration [5],
environmental monitoring [6], underwater [7], and even child
development [8].
This paper presents one such new design (Fig.1), but in doing
so, also attempts to quantify its performance to similar designs.
The design presented here is an omnidirectional tetrahedral
four-pendulum design that uses the position of the pendulums
to change the center of mass of the robot and thus create a
rolling torque. The robot is omnidirectional in that it can roll in
any direction instantly, unlike the more common single- or
double-pendulum designs. Using pendulums, the robot is
Figure 1. Sketch of four-pendulum mechanism
R
-R
-R
R
0
0
R
-R
0
+z
x
y
z
xz
+y
-y
rp
rm
θϕ
2 Copyright © 2015 by ASME
shown to be theoretically more agile than other existing
tetrahedral designs but more eccentric in its rotational inertia,
which may lead to more wobbling without proper control
schemes. Specifically, the new design is compared to similar
designs in terms of directionality, torque arm, inertia
eccentricity, and inertia magnitude.
RELATED WORK
Spherical robots can be roughly categorized by their
propulsion mechanism. One type of propulsion mechanism
uses conservation of angular momentum to induce rolling of a
rigid outer shell. These robots spin a heavy internal flywheel
(or flywheels), and then tilt it to produce shell rolling. For
example, Brown and Xu [9] built disk-on-edge robot that
moved via an internal gyroscope. Bhattacharya and Agrawal
[10] present a robot using two perpendicular rotors, while Shu
et al. [11] present a single rotor also acting as a liftable
pendulum. Joshi and Banavar [12] use four rotors that allow
for rolling in any direction. The main advantage of this type of
design is that it can produce torques equivalent to moving the
robot’s center of mass outside of its shell – more torque than
center-of-mass mechanism discussed below. One complication
of this design is that the resulting precession torque can move
the robot in unintended directions. The use of a rotor pair has
also been used by Schroll [13] to enhance a center-of-mass
design.
There are also examples of spherical robots that move by
distorting or transforming the shell itself. These designs are all
distinct and this vein of research is relatively unexplored. For
example, separate designs by Artusi et al. [14] and Wait et al.
[15] deform panels on the robot’s shell by dielectric actuators
and air bladders, respectively. Sugiyama et al. [16] propose a
wheel or shell made out of shape-memory alloys.
The most common propulsion scheme, and perhaps the
simplest to implement, is to change the robots internal center of
mass to induce a torque and thus rolling of a rigid outer shell.
Within this scheme, there are various methods changing that
center of mass, as discussed below. The main advantage of this
design is its actuation simplicity. However, the main limitation
is the torque is restricted by the envelope of possible center of
masses, as discussed below.
The first center-of-mass method in the literature is that of
an internal wheel or car mechanisms – often explained as a
robotic hamster in its wheel. The wheel or car rolls internally in
one direction, changing the robots center of mass, and inducing
rolling of the ball in that direction. For example, Halme et al.
[17] use a spring-loaded internal unicycle drive unit to shift the
center of mass. Bicchi and colleagues [18], and later Alves and
Dias [19], use a small car as the drive unit. From the external
(visible) point of view, these robots are omnidirectional in that
the shell can roll in any direction. However, they are not
instantaneously omnidirectional – the wheel or car must steer
first before being able to move in a sideways direction. They
also suffer from the typical grip-wear tradeoff of friction drives:
increasing the normal force on the drive wheels reduces drive
slip but also greatly increases steering friction and wear.
The most common method is to use a single two-degree-
of-freedom pendulum, here called the “single-pendulum”
design. Raising the pendulum causes the robot to roll forward
or backward, and tilting the pendulum while raising it causes
the robot to steer left or right while rolling. For example
Michaud and Caron [20] introduce Roball that is re-prototyped
as a toy for child development studies. Seeman et al [4] apply
the GroundBot robot to security tasks. Schroll [13] analyzes
the dynamics of a single-pendulum spherical robot that also
includes conservation of angular momentum enhancement.
Zheng et al. [21] analyze the BHQ-1 robot. Park et al [22]
proposes a control scheme for a single-pendulum robot.
Likewise, Pokhrel et al. [23] introduces SpheRobot, and Hogan
et al.i [5] discuss a wind-blown spherical robot that also uses a
hanging payload as a single pendulum. Recently, several
designs have emerged with two smaller hanging pendulums
that allow for stick-slip spin-in-place about the vertical axis.
Example designs are proposed and analyzed by Ghanbari et al.
[24], Yoon et al. [25], and Wang and Zhao [26].
These single-pendulum designs are more “wheel” than
“ball” because they all have a main axis/axle that the robot
must roll about. Thus they are not omnidirectional – when
facing forwards, they cannot purely roll sideways. However,
the single pendulum means that all of the dynamic mass of the
robot can be moved very close to the outer shell, maximizing
the torque created for a center-of-mass design.
A third way of moving the center of mass is by distributed
masses sliding along radial, such as tetrahedral, spokes. By
sliding the (four) masses as much to one side as possible, the
robot rolls in that direction – here, this type of design is called
“four-sliders”. Tomik et al. [27] first presented this design
along with a geometric motion planner. Javadi and Mojabi [28]
present a similar design with a simulation. The main advantage
of this design is that it is omnidirectional – the robot can move
in any direction instantly. It does not need to steer either via
spin or via rolling in a small arc. However, the distributed
masses means that some of the dynamic mass is not ideally
located for torque generation. For example, if attempting to
roll forward, one mass may be located on a spoke aimed
backwards and thus not able to apply a positive torque. (Even
worse, it may be applying a negative torque.)
A proposed but understudied design [29] uses cables and
pulleys to move a single mass about the interior of the robot.
This design allows for omnidirectional travel and maximized
center-of-mass torque, and thus has a lot of potential, but is not
heavily explored yet. Here, this design is called the “cable
mass” design.
Amid and in addition to the mechanism literature, there are
several veins of ball-on-plane dynamics applied to spherical
robots, motion planning, and spherical robot control research.
Besides such work included in the already cited papers – or by
similar papers by cited authors – is work done by Der et al.
[30], Qiang et al. [31], Peng et al. [32], Tao et al. [33], Ivanova
and Pivovarova [34], and Karavaev and Kilin [35]. The
3 Copyright © 2015 by ASME
dynamics and control of the four-pendulum mechanism
presented here are complicated (as they are for other spherical
robots), but they are not discussed in detail in this paper.
FOUR PENDULUM DESIGN
The spherical robot presented here is a center-of-mass
design that uses four pendulum masses in a tetrahedral
arrangement. Figure 1 shows a sketch of the (ideal) mechanism,
where R is the sphere’s radius. Each pendulum rotates about a
tetrahedral direction – changing the angle of the pendulums
changes the robot’s center of mass. This mechanism can create
rolling torque in any direction instantly, and in practice, the
pendulums are rotating in full circles as the robot rolls in any
one direction.
The main benefit of this design is its omnidirectionality,
with larger torque possible than previous tetrahedral designs.
The four pendulum design can roll in any direction
instantaneously, and can shift its center of mass farther forward
than the four-slider design, as will be shown here.
Figure 2 shows pictures of a physical prototype. The
robot’s electronics and battery are housed in and around central
rapid-prototyped tetrahedral box, and the four motors are
mounted facing out from that box. The box is centered in the
shell using four spokes that reach through gaps between
pendulum arcs, and the shell is a 33-cm-diameter plastic ball.
Figure 2. Pictures of the prototype four-pendulum
spherical robot.
The robot uses the Maple r5 as its processor. It is remote
controlled using a Futaba 75 MHz RF transmitter and receiver,
and carries one battery for the motors (11V 1500mAh LiPo)
and a smaller one for the electronics (3.7V 850mAh LiPo). The
motors are Pololu 12V brushed DC with a 50:1 reduction and
64 CPR encoders for position feedback. The robot senses
orientation via an onboard orientation sensor (CHR-UM6) that
processes accelerometer, rate gyro, and magnetic sensor data.
Given the desired torque magnitude and direction from the RF
data, the robot computes the desired center of mass location in
world coordinates, converts that position to robot and then
pendulum coordinates, and moves the pendulums to achieve
that center.
The tetrahedral transformations are as follows (see labels
in Fig. 1). The direct angle between axes is the standard
tetrahedral angle,
φ = cos−1
−
1
3
"
#
$
%
&
' ≈109.5° , (1)
and the z-rotation from “xz” pendulum to “+y” pendulum is
ϕ =120° . (2)
Let the pendulums be identical with length rp rotating on motor
axes at tetrahedral distances of rm. Given pendulum i's rotation
angle, θi, the position of that pendulum in robot coordinates, xi,
is
xi = Ri
rp cosθ
rp sinθ
rm
!
"
#
#
#
#
$
%
&
&
&
&
, (3)
where Ri is the rotation matrix for that pendulum. Here,
R+z =
1 0 0
0 1 0
0 0 1
!
"
#
#
#
$
%
&
&
&
, (4)
Rxz =
cosφ 0 sinφ
0 1 0
−sinφ 0 cosφ
"
#
$
$
$
$
%
&
'
'
'
'
, (5)
R+y =
cosϕ −sinϕ 0
sinϕ cosϕ 0
0 0 1
"
#
$
$
$
$
%
&
'
'
'
'
• Rxz , (6)
R−y =
cosϕ sinϕ 0
−sinϕ cosϕ 0
0 0 1
"
#
$
$
$
$
%
&
'
'
'
'
• Rxz . (7)
For maximum rolling torque, the pendulum length should
be as large as possible without collisions between pendulums.
That is, larger rp results in larger torque. For this design, the
maximum rp without collision is when two pendulum arcs
touch – by trigonometry, this occurs when
rp = Rsin
φ
2
!
"
#
$
%
& ≈ 0.82R
rm = Rcos
φ
2
!
"
#
$
%
& ≈ 0.58R
. (8)
Simulations confirm that these dimensions result in the largest
possible torque.
The following sections compare this mechanism to similar
designs with a summary shown in Table 1. The comparisons
are made for the ideal, theoretical designs (no static mass, all
point-mass dynamic mass). Such a comparison is useful
separate from the pragmatic comparison of costs, realistic
masses and dimensions, and ease of construction or control.
4 Copyright © 2015 by ASME
Design Single
Pendulum
Four
Sliders
Cable
Mass
Four
Pendulums
Directionality Steered Omni Omni Omni
Torque arm, rT <100% <29% <100% <67%
Eccentricity, e 100% 0-20% 0 or
100%
36-100%
Radius of
Gyration, k
100% 0-178% 0-100% 87-216%
Table 1. Comparison of Several Theoretical Center-of-
Mass Designs
DIRECTIONALITY COMPARISON
This section compares the four-pendulum design to
existing center-of-mass designs in terms of what is called
“directionality”. This term is not meant as a technical term and
applies to the robot’s shell on the (floor) plane.
A sphere on a plane has five degrees-of-freedom: two
translational degrees corresponding to the location of the
sphere’s center, and three rotational degrees corresponding to
the sphere’s orientation at the point. The ideal spherical robot
would be able to change any of these degrees instantaneously
(and thus be in the technical sense, holonomic).
However, a rolling sphere is non-holonomic in that it has
no-slip conditions and must roll to change its orientation (and
position). An ideal rolling spherical robot would be able to roll
in any direction instantaneously without steering – similar to a
sports ball rolling on a floor or field. Such omnidirectionality
has obvious advantages when navigating complex
environments. In this sense, the four-pendulum, sliding-
masses, and cable-mass designs are omnidirectional, while the
more prevalent single-pendulum designs are not.
TORQUE ARM COMPARISON
The second comparison to be made between the four-
pendulum design and existing center-of-mass designs is in
terms of its theoretical normalized torque arm. Here, the
torque arm, rT is defined as the effective torque arm of the
dynamic center of mass envelope, as a percentage of the sphere
radius:
rT =
horizontal distance to dynamic CoM
sphere radius
•100%
=
xCoM
2
+ yCoM
2
R
•100%
. (8)
Thus, normalized torque can be 0 to 100% for center-of-mass
designs, but larger than that for conservation-of-angular-
momentum designs.
Each robot implementation has some “static” mass
consisting of its components that do not move to change the
center of mass and thus the drive torque – its shell, internal
framework, and possibly electronics, sensors, battery, and
motors. It also has some “dynamic” mass that can be moved to
change its center of mass – the pendulums or sliding masses,
which may include some or all of the electronics, sensors,
battery, and motors. A robot-building design goal is to
minimize the static mass while maximizing the dynamic mass,
to allow for the maximum possible torque.
A theoretical-mechanism design goal is to create as large a
torque arm as possible from the dynamic mass, assuming the
static mass is negligible. In this sense, the single-pendulum and
cable-mass designs have the largest torque arm possible of
100% – if modeled as a point mass, the center of mass can be
moved anywhere in the sphere, out to the sphere radius. (In
application, the static mass decreases this to a more typical 50-
67% [13].)
The four-sliding-mass design, while omnidirectional, has a
much smaller torque arm. If the masses are equal, then
simulation shows that its center of mass envelope for all
possible slider configurations is a polyhedron with a maximum
torque arm of 29%. The maximum torque arm also varies
based on the direction – in some directions, it is only 20%.
The four-pendulum design is also omnidirectional, and has
a larger torque arm than the four-slider, but less than the single-
pendulum and cable-mass designs. If the masses are equal,
then its center of mass can be calculated as the sum of the four
xi’s (Eqn. (3)) and the envelope of possible locations looks like
a sphere with eight truncations (see Fig. 3). The figure is
generated by simulating all possible configurations of the four
pendulums, and calculating the resulting dynamic center of
mass, but then plotting only the outer envelope. The maximum
torque arm in any direction ranges from 56% to 67%. That is,
the four-pendulum can create a torque equivalent to moving its
dynamic center of mass approximately two-thirds along its
sphere radius.
Figure 3. Four-pendulum theoretical torque-arm
envelope.
ECCENTRICITY COMPARISON
In addition, this section compares the four-pendulum
design to existing center-of-mass designs in terms of the
eccentricity, e, of the robot’s rotational moment of inertia, I,
about its center.
−0.5
0
0.5
−0.5
0
0.5
−0.5
0
0.5
z(normalized)
x (normalized)
y (normalized)
5 Copyright © 2015 by ASME
A general three-dimensional body has a moment of inertia
ellipsoid representing its angular inertia in all directions. The
axes of the inertia ellipsoid are called the principal axes, and
their magnitudes are the principal moments of inertia. The
largest inertia, Imajor, about any axis corresponds to the
ellipsoid’s semi-major axis, and the smallest, Iminor, corresponds
to the ellipsoid’s semi-minor axis.
Let Ixx, Iyy, and Izz be the moments of inertia about the x, y,
and z axes, and Ixy, Iyz, and Ixz be the corresponding products of
inertia. Then, the inertia matrix representing the inertia
ellipsoid is
I =
Ixx −Ixy −Ixz
−Ixy Iyy −Iyz
−Ixz −Iyz Izz
"
#
$
$
$
$
%
&
'
'
'
'
. (9)
The ellipsoid’s principal axes and values can be found from the
standard eigenvector problem
Av = λv (10)
where generic A is now the inertia matrix I; the eigenvalues, λ,
are the magnitude of the principal inertias/axes, and the
corresponding eigenvectors, v, are the cosines of the principal
axes with respect to the x, y, and z axes.
One of the well-known challenges in spherical robots is
that the robots tend to wobble as they roll. Angular momentum
( H = Iω ) is conserved, but because the robot’s moment of
inertia (I) is non-homogeneous (i.e., an ellipsoid and not a
sphere), the angular momentum (ω) also varies. The result is
similar to an American football that wobbles outside of a
perfect spiral.
Thus, having a homogeneous inertia (i.e., a spherical
inertia ellipsoid) is advantageous in spherical robots because it
reduces wobble. The more non-homogeneous, or “eccentric”,
the inertia ellipsoid is, the more the wobble. Here, eccentricity
is defined in the standard mathematical sense (also called first
eccentricity) but as a percentage:
e = 1−
Iminor
2
Imajor
2
•100%. (11)
In application, any static mass usually decreases eccentricity
because the overall mass is more distributed than the pendulum
masses.
A single-mass (pendulum or cable-controlled) design with
a point-mass bob has an inertia ellipse collapsed into a line – its
eccentricity is always 100% unless, in the cable-mass design,
the mass is in the center of the sphere and the eccentricity has
approached 0%. Thus, in any torque situation, these designs
have the potential for large wobble if control does not
compensate.
In terms of eccentricity, the four-slider design is much
better. Like the cable-mass design, when all four point masses
are at the center of the sphere, the eccentricity is 0%.
Otherwise, simulation shows it varies with a maximum of only
20% – the distributed-mass four-slider design is much less
prone to wobble than the single-pendulum or cable-mass
designs. This effect makes sense – distributing the dynamic
mass around the sphere homogenizing the ellipsoid, decreasing
the eccentricity.
The four-pendulum design is in-between the existing
designs. At no point can the four pendulums be moved into the
middle like the cable-mass or four-slider design, so the
eccentricity is never 0; it’s minimum is 36% (although while
still applying a torque). On the other hand, when the four
pendulums are paired up on opposite sides of the spheres, the
eccentricity is 100% – balanced about the center, but collapsed
into a line. A simulation shows that for all pendulum locations,
the eccentricity ranges for 36 to 100% with an average of 84%.
Thus, the four-pendulum design is more prone to wobble than
the four-slider design, but less than the single-mass designs.
GYRATION COMPARISON
The last comparison to be made between the designs is the
radius of gyration – a standard sense of the magnitude of a
moment of inertia. The normalized radius of gyration, k, is
defined here as percentage of the sphere’s radius:
k =
I
m
•
100%
R
. (12)
It can be visualized as the effective radius to a point mass that
results in the same moment of inertia. A larger radius of
gyration is equivalent to a larger moment of inertia, which can
be beneficial when maintaining angular speed, or detrimental
when trying to change angular speed. Usually, a design goal is
to minimize inertia to allow for a more nimble robot.
This is an important difference from minimizing
eccentricity. A homogenous sphere has an eccentricity of 0%
and a non-zero radius of gyration; the same mass as a point
mass can be placed at the same radius of gyration but will now
have an eccentricity of 100%.
For the single-pendulum design, the maximum theoretical
k is always 100% because there is only one point mass, at the
sphere’s radius. For the cable-mass design, the maximum k
ranges from 0 to 100%, linearly related to the radius to the
mass. For the four-slider design, the maximum k varies from 0
(when all masses are pulled in) to 178% (when all masses are
extended). For the four-pendulum design, the maximum k
varies from 81% to 218% – this design has a large moment of
inertia because of the masses cannot collapse into the center.
Thus the ranking of the four designs with regards to radius of
gyration depends on their configuration, but the cable-mass and
four-slider designs have the best possible configurations.
PROTOTYPE NUMBERS
The physical prototype was designed and built as a proof-
of-concept of the four-pendulum driving mechanism, and as an
initial testbed for control schemes. As such, it is a success.
The robot successfully rolls omnidirectionally as commanded
through the RF communication, and maintains (visually) proper
heading.
It should be noted that the comparisons above are for the
theoretical mechanism, and are dampened by the physical
application.
6 Copyright © 2015 by ASME
For example, the torque arm as calculated (56-67%) is for
the ideal mechanism; the physical prototype has a much smaller
torque arm. This prototype version is not optimized for
dynamic mass over static mass. It has a static mass of shell,
batteries, motors, and electronics totaling 2.6kg, with a
dynamic mass of 1.6kg – in Chase and Pandya’s [1] terms, its
power factor is 0.615. For a 33-cm-diameter sphere, the four
pendulums have ideal lengths of around 13.5 cm; because of
clearances and non-point-masses, the pendulum have effective
lengths of around 10 cm. The effective torque arm of the
prototype is closer to 20%. Initial tests with this prototype saw
it accelerate with 0.6 m/s2
up to a maximum speed of 0.7 m/s.
On the other hand, the eccentricity of the prototype is
better than theoretical. The inertia of the static masses means
that the inertia ellipse is less dependent on the pendulum
locations. The theoretical eccentricity ranges from 36% to
100% with an average of 84%; the prototype’s eccentricity
ranges from 30% to 92% with an average of 73%.
The radius of gyration also increases with the static mass.
For the prototype, it increases to 155-250%.
CONCLUSION
The four-pendulum omnidirectional spherical mobile robot
design has been presented and discussed. Compared to previous
center-of-mass designs, the four-pendulum design has large
directionality, medium torque arm, medium eccentricity, but
large radius of gyration. It is a viable spherical robot design.
A physical prototype has been built and tested. It
successfully rolls using the pendulums to change the center of
mass of the robot. The prototype is successful even with a
decreased torque arm and increased radius of gyration, and is
less prone to wobble than its theoretical equivalent.
Future work includes rigorous dynamical analysis, robust
control implementations, and a more optimized prototype. The
comparisons also suggest that the cable-mass mechanism is a
beneficial design and should be analyzed further.
ACKNOWLEDGMENTS
This work was funded in part by the Central Michigan
University FRCE Grant #48868.
REFERENCES
[1] Chase, R., and Pandya, A., 2012. “A review of active
mechanical driving principles of spherical robots”. Robotics, 1,
pp. 3–23.
[2] Crossley, V. A., 2006. A literature review of the design of
spherical rolling robots.
[3] Armour, R. H., and Vincent, J. F. V., 2006. “Rolling in
nature and robotics: A review”. Journal of Bionic Engineering,
3(4), 12, pp. 195–208.
[4] Seeman, M., Broxvall, M., Saffiotti, A., and Wide, P., 2006.
“An autonomous spherical robot for security tasks”. In IEEE
Conference on Computational Intelligence for Homeland
Security and Personal Safety.
[5] Hogan, F. R., Forbes, J. R., and Barfoot, T. D., 2014.
“Rolling stability of a power-generating tumbleweed rover”.
Journal of Spacecraft and Rockets, 51(6), pp. 1895–1906.
[6] Hernandez, J. D., Barrientos, J., del Cerro, J., Barrientos,
A., and Sanz, D., 2013. “Moisture measurement in crops using
spherical robots”. Industrial Robot: An International Journal,
40(1), pp. 59–66.
[7] Li, M., Guo, S., Hirata, H., and Ishihara, H., 2015. “Design
and performance evaluation of an amphibious spherical robot”.
Robotics and Autonomous Systems, 64, pp. 21–34.
[8] Michaud, F., Laplante, J.-F., Larouche, H., Duquette, A.,
Caron, S., Letourneau, D., and Masson, P., 2005. “Autonomous
spherical mobile robot for child-development studies”. IEEE
Transactions on Systems, Man, and Cybernetics, 35(4), July,
pp. 471–9.
[9] Brown, H. B. J., and Xu, Y., 1997. “A single-wheel
gyroscopically stabiliszed robot”. IEEE Robotics and
Automation Magazine, September.
[10] Bhattacharya, S., and Agrawal, S. K., 2000. “Spherical
rolling robot: A design and motion planning studies”. IEEE
Transactions of Robotics and Automation, 16(6), pp. 835–9.
[11] Shu, G., Zhan, Q., and Cai, Y., 2009. “Motion control of
spherical robot based on conservation of angular momentum”.
In ICMA International Conference on Mechatronics and
Automation, pp. 599–604.
[12] Joshi, V. A., Banavar, R. N., and Hippalgaonkar, R., 2010.
“Design and analysis of a spherical mobile robot”. Mechanism
and Machine Theory, 45(2), 2, pp. 130–136.
[13] Schroll, G. C., 2010. “Dynamic model of a spherical robot
from first principles”. Master’s thesis, Colorado State
University, Summer.
[14] Artusi, M., Potz, M., Aristizabal, J., Menon, C., Cocuzza,
S., and Debei, S., 2011. “Electroactive elastomeric actuators for
the implementation of a deformable spherical rover”.
IEEE/ASME Transactions on Mechatronics, 16(1), pp. 50–57.
[15] Wait, K. W., Jackson, P. J., and Smoot, L. S., 2010. “Self
locomotion of a spherical rolling robot using a novel
deformable pneumatic method”. In IEEE International
Conference on Robotics and Automation.
[16] Sugiyama, Y., Shiotsu, A., Yamanaka, M., and Hirai, S.,
2005. “Circular/spherical robots for crawling and jumping”. In
IEEE International Conference on Robotics and Automation.
[17] Halme, A., Torsten, and Wang, Y., 1996. “Motion control
of a spherical mobile robot”. In 4th International Workshop on
Advanced Motion Control, pp. 259–64.
[18] Bicchi, A., Balluchi, A., Prattichizzo, D., and Gorelli, A.,
1997. “Introducing the Sphericle: an experimental testbed for
research and teaching in nonholonomy”. In IEEE International
Conference on Robotics and Automation, pp. 2620–5.
[19] Alves, J., and Dias, J., 2003. “Design and control of a
spherical mobile robot”. Journal of Systems and Control
Engineering, 217, September, pp. 457–467.
[20] Michaud, F., and Caron, S., 2002. “Roball, the rolling
robot”. Autonomous Robots, 12(2), pp. 211–222.
[21] Zheng, M., Qiang, Z., Jinkun, L., and Yao, C., 2011.
“Control of a spherical robot: Path following based on
7 Copyright © 2015 by ASME
nonholonomic kinematics and dynamics”. Chinese Journal of
Aeronautics, 24(3), pp. 337–345.
[22] Gye-Do Park, K.-H. K., Ba Da, Y., and Lee, J.-M., 2011.
“Roll control of spherical robot using the IMU”. In 3rd
SPENALO International Symposium.
[23] Pokhrel, D., Luitel, N. R., Das, S., and Ray, D. N., 2013.
“Design and development of a spherical robot (SpheRobot)”. In
1st International and 16th National Conference on Machines
and Mechanisms.
[24] Ghanbari, A., Mahboubi, S., and Fakhrabadi, M., 2010.
“Design, dynamic modeling and simulation of a spherical
mobile robot with a novel motion mechanism”. In
Mechatronics and Embedded Systems and Applications
(MESA), IEEE/ASME International Conference on, pp. 434–
439.
[25] Yoon, J.-C., Ahn, S.-S., and Lee, Y.-J., 2011. “Spherical
robot with new type of two-pendulum driving mechanism”. In
Intelligent Engineering Systems (INES), 15th IEEE
International Conference on, IEEE, pp. 275–279.
[26] Wang, L., and Zhao, B., 2015. “Dynamic modeling and
control strategy for turning in place motion of a two-coaxial
pendulums driven spherical inspector based on stick–slip
principle”. Mechanism and Machine Theory, 83, pp. 69–80.
[27] Tomik, F., Nudehi, S., Flynn, L. L., and Mukherjee, R.,
2012. “Design, fabrication and control of SpheRobot: A
spherical mobile robot”. Journal of Intelligent Robot Systems,
67(2), pp. 117–131.
[28] Javadi A., A. H., and Mojabi, P., 2004. “Introducing glory:
A novel strategy for an omnidirectional spherical rolling robot”.
Journal of Dynamic Systems, Measurement, and Control,
126(3), pp. 678–83.
[29] Lux, J., 2005. Alternative way of shifting mass to move a
spherical robot. Tech. rep., NASA’s Jet Propulsion Laboratory,
June.
[30] Der, R., Martius, G., and Hesse, F., 2006. “Let it roll —
emerging sensorimotor coordination in a spherical robot”. In
International Conference on the Simulation and Synthesis of
Living Systems, MIT Press, pp. 192–198.
[31] Qiang, Z., Zengbo, L., and Yao, C., 2008. “A back-
stepping based trajectory tracking controller for a non-chained
nonholonomic spherical robot”. Chinese Journal of
Aeronautics, 21(5), pp. 472–480.
[32] Peng, Y.-F., Chiu, C.-H., Tsai, W.-R., and Chou, M.-H.,
2009. “Design of an omni-directional spherical robot: Using
fuzzy control”. In The International Multiconference of
Engineers and Computer Scientists, Vol. 1, pp. 18–20.
[33] Tao, Y., Hanxu, S., and Qingxuan, J., 2012. “Variable
structure control of pendulum-driven spherical mobile robots”.
In 3rd International Conference on Computer and Electrical
Engineering (ICCEE 2010 no. 2).
[34] Ivanova, T., and Pivovarova, E., 2013. “Dynamics and
control of a spherical robot with an axisymmetric pendulum
actuator”. Nonlin. Dyn. Mob. Robot, 1(1), pp. 71–85.
[35] Karavaev, Y. L., and Kilin, A. A., 2015. “The dynamics
and control of a spherical robot with an internal omniwheel
platform”. Regular and Chaotic Dynamics, 20(2), pp. 134–152.

More Related Content

Viewers also liked

Попередня програма щорічної конференції
Попередня програма щорічної конференціїПопередня програма щорічної конференції
Попередня програма щорічної конференції
USAID LEV
 
Recent Trends in Income Inequality in Finland
Recent Trends in Income Inequality in FinlandRecent Trends in Income Inequality in Finland
Recent Trends in Income Inequality in Finland
Palkansaajien tutkimuslaitos
 
Batur caldera premier
Batur caldera premierBatur caldera premier
Batur caldera premier
Jero Wijaya
 
D. helms resume 2015
D. helms resume 2015D. helms resume 2015
D. helms resume 2015
Danny Helms Jr.
 
Vihaan kaushik
Vihaan kaushikVihaan kaushik
Vihaan kaushik
Renaissance123
 
Lallder_Mar2016skillresume
Lallder_Mar2016skillresumeLallder_Mar2016skillresume
Lallder_Mar2016skillresume
Linda Allder
 
Resumen
ResumenResumen
Resumen
Marii Montoya
 
Evidencias científicas de la creación bíblica
Evidencias científicas de la creación bíblicaEvidencias científicas de la creación bíblica
Evidencias científicas de la creación bíblica
SilverWolf Aliaga
 

Viewers also liked (11)

Degree
DegreeDegree
Degree
 
Попередня програма щорічної конференції
Попередня програма щорічної конференціїПопередня програма щорічної конференції
Попередня програма щорічної конференції
 
Recent Trends in Income Inequality in Finland
Recent Trends in Income Inequality in FinlandRecent Trends in Income Inequality in Finland
Recent Trends in Income Inequality in Finland
 
Batur caldera premier
Batur caldera premierBatur caldera premier
Batur caldera premier
 
D. helms resume 2015
D. helms resume 2015D. helms resume 2015
D. helms resume 2015
 
Vihaan kaushik
Vihaan kaushikVihaan kaushik
Vihaan kaushik
 
Lallder_Mar2016skillresume
Lallder_Mar2016skillresumeLallder_Mar2016skillresume
Lallder_Mar2016skillresume
 
Resumen
ResumenResumen
Resumen
 
finale_mus
finale_musfinale_mus
finale_mus
 
Evidencias científicas de la creación bíblica
Evidencias científicas de la creación bíblicaEvidencias científicas de la creación bíblica
Evidencias científicas de la creación bíblica
 
Портфоліо
ПортфоліоПортфоліо
Портфоліо
 

Similar to fourpend2015_v05

D012222333
D012222333D012222333
D012222333
IOSR Journals
 
Vehicle-applicable robots controlled byMobile
Vehicle-applicable robots controlled byMobileVehicle-applicable robots controlled byMobile
Vehicle-applicable robots controlled byMobile
IRJET Journal
 
A Review of Shortcomings of Driving principle for Spherical Drive Systems
A Review of Shortcomings of Driving principle for Spherical Drive SystemsA Review of Shortcomings of Driving principle for Spherical Drive Systems
A Review of Shortcomings of Driving principle for Spherical Drive Systems
paperpublications3
 
Design of Quadruped Walking Robot with Spherical Shell
Design of Quadruped Walking Robot with Spherical ShellDesign of Quadruped Walking Robot with Spherical Shell
Design of Quadruped Walking Robot with Spherical Shell
drboon
 
Design, Modification and Manufacturing of a Stair Climbing Vehicle
Design, Modification and Manufacturing of a Stair Climbing VehicleDesign, Modification and Manufacturing of a Stair Climbing Vehicle
Design, Modification and Manufacturing of a Stair Climbing Vehicle
Raian Nur Islam
 
SIX DEGREES OF FREEDOM INTEGRATION DEVICE
SIX DEGREES OF FREEDOM INTEGRATION DEVICESIX DEGREES OF FREEDOM INTEGRATION DEVICE
SIX DEGREES OF FREEDOM INTEGRATION DEVICE
IAEME Publication
 
IRJET- Design and Fabrication of Rocker Bogie Mechanism using Solar Energy
IRJET- Design and Fabrication of Rocker Bogie Mechanism using Solar EnergyIRJET- Design and Fabrication of Rocker Bogie Mechanism using Solar Energy
IRJET- Design and Fabrication of Rocker Bogie Mechanism using Solar Energy
IRJET Journal
 
Km229 automobile diwheel motorized vehicle
Km229 automobile diwheel motorized vehicleKm229 automobile diwheel motorized vehicle
Km229 automobile diwheel motorized vehicle
1000kv technologies
 
IRJET- Review on Rover with Rocker-Bogie Linkage Mounted with Ultrasonic Sens...
IRJET- Review on Rover with Rocker-Bogie Linkage Mounted with Ultrasonic Sens...IRJET- Review on Rover with Rocker-Bogie Linkage Mounted with Ultrasonic Sens...
IRJET- Review on Rover with Rocker-Bogie Linkage Mounted with Ultrasonic Sens...
IRJET Journal
 
Mobile Robot Vechiles
Mobile Robot VechilesMobile Robot Vechiles
Mobile Robot Vechiles
cairo university
 
Gyroscopic effect on bearings
Gyroscopic effect on bearingsGyroscopic effect on bearings
Gyroscopic effect on bearings
Akansha Jha
 
A Dynamic Single Actuator Vertical Climbing Robot
A Dynamic Single Actuator Vertical Climbing RobotA Dynamic Single Actuator Vertical Climbing Robot
A Dynamic Single Actuator Vertical Climbing Robot
Tina Gabel
 
IRJET- Design and Fabrication of Multi Legged Robot
IRJET-  	  Design and Fabrication of Multi Legged RobotIRJET-  	  Design and Fabrication of Multi Legged Robot
IRJET- Design and Fabrication of Multi Legged Robot
IRJET Journal
 
Theory of machines
Theory of machinesTheory of machines
Theory of machines
Sourabh Kumar
 
1 s2.0-s0094114 x15002402-main
1 s2.0-s0094114 x15002402-main1 s2.0-s0094114 x15002402-main
1 s2.0-s0094114 x15002402-main
Mesfin Demise
 
Slider Crank Mechanism for Four bar linkage
Slider Crank Mechanism for Four bar linkageSlider Crank Mechanism for Four bar linkage
Slider Crank Mechanism for Four bar linkage
ijsrd.com
 
Modeling, Simulation, and Optimal Control for Two-Wheeled Self-Balancing Robot
Modeling, Simulation, and Optimal Control for Two-Wheeled Self-Balancing Robot Modeling, Simulation, and Optimal Control for Two-Wheeled Self-Balancing Robot
Modeling, Simulation, and Optimal Control for Two-Wheeled Self-Balancing Robot
IJECEIAES
 
Geometrical Analysis and Design of Tension-Actuated Ackermann Steering System...
Geometrical Analysis and Design of Tension-Actuated Ackermann Steering System...Geometrical Analysis and Design of Tension-Actuated Ackermann Steering System...
Geometrical Analysis and Design of Tension-Actuated Ackermann Steering System...
Scientific Review SR
 
04 mobile robot vehicles
04 mobile robot vehicles04 mobile robot vehicles
04 mobile robot vehicles
cairo university
 
Design and analysis of x y- positioning stage based on redundant parallel li...
Design and analysis of x y- positioning stage based on redundant parallel li...Design and analysis of x y- positioning stage based on redundant parallel li...
Design and analysis of x y- positioning stage based on redundant parallel li...
eSAT Journals
 

Similar to fourpend2015_v05 (20)

D012222333
D012222333D012222333
D012222333
 
Vehicle-applicable robots controlled byMobile
Vehicle-applicable robots controlled byMobileVehicle-applicable robots controlled byMobile
Vehicle-applicable robots controlled byMobile
 
A Review of Shortcomings of Driving principle for Spherical Drive Systems
A Review of Shortcomings of Driving principle for Spherical Drive SystemsA Review of Shortcomings of Driving principle for Spherical Drive Systems
A Review of Shortcomings of Driving principle for Spherical Drive Systems
 
Design of Quadruped Walking Robot with Spherical Shell
Design of Quadruped Walking Robot with Spherical ShellDesign of Quadruped Walking Robot with Spherical Shell
Design of Quadruped Walking Robot with Spherical Shell
 
Design, Modification and Manufacturing of a Stair Climbing Vehicle
Design, Modification and Manufacturing of a Stair Climbing VehicleDesign, Modification and Manufacturing of a Stair Climbing Vehicle
Design, Modification and Manufacturing of a Stair Climbing Vehicle
 
SIX DEGREES OF FREEDOM INTEGRATION DEVICE
SIX DEGREES OF FREEDOM INTEGRATION DEVICESIX DEGREES OF FREEDOM INTEGRATION DEVICE
SIX DEGREES OF FREEDOM INTEGRATION DEVICE
 
IRJET- Design and Fabrication of Rocker Bogie Mechanism using Solar Energy
IRJET- Design and Fabrication of Rocker Bogie Mechanism using Solar EnergyIRJET- Design and Fabrication of Rocker Bogie Mechanism using Solar Energy
IRJET- Design and Fabrication of Rocker Bogie Mechanism using Solar Energy
 
Km229 automobile diwheel motorized vehicle
Km229 automobile diwheel motorized vehicleKm229 automobile diwheel motorized vehicle
Km229 automobile diwheel motorized vehicle
 
IRJET- Review on Rover with Rocker-Bogie Linkage Mounted with Ultrasonic Sens...
IRJET- Review on Rover with Rocker-Bogie Linkage Mounted with Ultrasonic Sens...IRJET- Review on Rover with Rocker-Bogie Linkage Mounted with Ultrasonic Sens...
IRJET- Review on Rover with Rocker-Bogie Linkage Mounted with Ultrasonic Sens...
 
Mobile Robot Vechiles
Mobile Robot VechilesMobile Robot Vechiles
Mobile Robot Vechiles
 
Gyroscopic effect on bearings
Gyroscopic effect on bearingsGyroscopic effect on bearings
Gyroscopic effect on bearings
 
A Dynamic Single Actuator Vertical Climbing Robot
A Dynamic Single Actuator Vertical Climbing RobotA Dynamic Single Actuator Vertical Climbing Robot
A Dynamic Single Actuator Vertical Climbing Robot
 
IRJET- Design and Fabrication of Multi Legged Robot
IRJET-  	  Design and Fabrication of Multi Legged RobotIRJET-  	  Design and Fabrication of Multi Legged Robot
IRJET- Design and Fabrication of Multi Legged Robot
 
Theory of machines
Theory of machinesTheory of machines
Theory of machines
 
1 s2.0-s0094114 x15002402-main
1 s2.0-s0094114 x15002402-main1 s2.0-s0094114 x15002402-main
1 s2.0-s0094114 x15002402-main
 
Slider Crank Mechanism for Four bar linkage
Slider Crank Mechanism for Four bar linkageSlider Crank Mechanism for Four bar linkage
Slider Crank Mechanism for Four bar linkage
 
Modeling, Simulation, and Optimal Control for Two-Wheeled Self-Balancing Robot
Modeling, Simulation, and Optimal Control for Two-Wheeled Self-Balancing Robot Modeling, Simulation, and Optimal Control for Two-Wheeled Self-Balancing Robot
Modeling, Simulation, and Optimal Control for Two-Wheeled Self-Balancing Robot
 
Geometrical Analysis and Design of Tension-Actuated Ackermann Steering System...
Geometrical Analysis and Design of Tension-Actuated Ackermann Steering System...Geometrical Analysis and Design of Tension-Actuated Ackermann Steering System...
Geometrical Analysis and Design of Tension-Actuated Ackermann Steering System...
 
04 mobile robot vehicles
04 mobile robot vehicles04 mobile robot vehicles
04 mobile robot vehicles
 
Design and analysis of x y- positioning stage based on redundant parallel li...
Design and analysis of x y- positioning stage based on redundant parallel li...Design and analysis of x y- positioning stage based on redundant parallel li...
Design and analysis of x y- positioning stage based on redundant parallel li...
 

fourpend2015_v05

  • 1. 1 Copyright © 2015 by ASME DRAFT Proceedings of the ASME 2015 International Mechanical Engineering Congress & Exposition IMECE2015 November 13-19, 2015, Houston, Texas IMECE2015-51551 A FOUR-PENDULUM OMNIDIRECTIONAL SPHERICAL ROBOT: DESIGN ANALYSIS AND COMPARISON Brian P. DeJong Central Michigan University Mount Pleasant, MI, USA Kumar Yelamarthi Central Michigan University Mount Pleasant, MI, USA Brad Bloxsom Central Michigan University Mount Pleasant, MI, USA ABSTRACT This paper presents a four-pendulum design for an omnidirectional spherical robot. The robot moves by the rotation of tetrahedrally-located pendulums to change the center of mass, and can do so instantaneously in any direction. In that respect, it is omnidirectional unlike the more common single- pendulum designs. In the paper, the theoretical design is discussed and contrasted to existing theoretical center-of-mass designs in terms of directionality, torque arm, eccentricity, and radius of gyration. The torque arm comparison is based on the envelope of the dynamically-controlled mass (i.e., how far in any direction can the center of mass be moved), while the eccentricity and gyration criteria are based on homogeneity and size of the mass moment of inertia (i.e., how much does the design “wobble” while rolling; how much inertia does it have). In these regards, the four-pendulum architecture is a strong design – better than some in one respect, better than the others in other respects. A prototype has been built and tested that successfully implements the four-pendulum propulsion, and its results are compared to theoretical ones. INTRODUCTION Spherical mobile robots are an interesting research area with subtly complex dynamics and a variety of mechanisms [1- 3]. Various propulsion schemes have been used, and since one scheme has not emerged clearly better than the rest, new designs are still being presented. The ball-shaped robots are intriguing because they have often have no externally-visible sources of propulsion and yet can navigate the non-holonomic rolling-ball-on-plane space. They have application in patrol and surveillance [4], extraterrestrial exploration [5], environmental monitoring [6], underwater [7], and even child development [8]. This paper presents one such new design (Fig.1), but in doing so, also attempts to quantify its performance to similar designs. The design presented here is an omnidirectional tetrahedral four-pendulum design that uses the position of the pendulums to change the center of mass of the robot and thus create a rolling torque. The robot is omnidirectional in that it can roll in any direction instantly, unlike the more common single- or double-pendulum designs. Using pendulums, the robot is Figure 1. Sketch of four-pendulum mechanism R -R -R R 0 0 R -R 0 +z x y z xz +y -y rp rm θϕ
  • 2. 2 Copyright © 2015 by ASME shown to be theoretically more agile than other existing tetrahedral designs but more eccentric in its rotational inertia, which may lead to more wobbling without proper control schemes. Specifically, the new design is compared to similar designs in terms of directionality, torque arm, inertia eccentricity, and inertia magnitude. RELATED WORK Spherical robots can be roughly categorized by their propulsion mechanism. One type of propulsion mechanism uses conservation of angular momentum to induce rolling of a rigid outer shell. These robots spin a heavy internal flywheel (or flywheels), and then tilt it to produce shell rolling. For example, Brown and Xu [9] built disk-on-edge robot that moved via an internal gyroscope. Bhattacharya and Agrawal [10] present a robot using two perpendicular rotors, while Shu et al. [11] present a single rotor also acting as a liftable pendulum. Joshi and Banavar [12] use four rotors that allow for rolling in any direction. The main advantage of this type of design is that it can produce torques equivalent to moving the robot’s center of mass outside of its shell – more torque than center-of-mass mechanism discussed below. One complication of this design is that the resulting precession torque can move the robot in unintended directions. The use of a rotor pair has also been used by Schroll [13] to enhance a center-of-mass design. There are also examples of spherical robots that move by distorting or transforming the shell itself. These designs are all distinct and this vein of research is relatively unexplored. For example, separate designs by Artusi et al. [14] and Wait et al. [15] deform panels on the robot’s shell by dielectric actuators and air bladders, respectively. Sugiyama et al. [16] propose a wheel or shell made out of shape-memory alloys. The most common propulsion scheme, and perhaps the simplest to implement, is to change the robots internal center of mass to induce a torque and thus rolling of a rigid outer shell. Within this scheme, there are various methods changing that center of mass, as discussed below. The main advantage of this design is its actuation simplicity. However, the main limitation is the torque is restricted by the envelope of possible center of masses, as discussed below. The first center-of-mass method in the literature is that of an internal wheel or car mechanisms – often explained as a robotic hamster in its wheel. The wheel or car rolls internally in one direction, changing the robots center of mass, and inducing rolling of the ball in that direction. For example, Halme et al. [17] use a spring-loaded internal unicycle drive unit to shift the center of mass. Bicchi and colleagues [18], and later Alves and Dias [19], use a small car as the drive unit. From the external (visible) point of view, these robots are omnidirectional in that the shell can roll in any direction. However, they are not instantaneously omnidirectional – the wheel or car must steer first before being able to move in a sideways direction. They also suffer from the typical grip-wear tradeoff of friction drives: increasing the normal force on the drive wheels reduces drive slip but also greatly increases steering friction and wear. The most common method is to use a single two-degree- of-freedom pendulum, here called the “single-pendulum” design. Raising the pendulum causes the robot to roll forward or backward, and tilting the pendulum while raising it causes the robot to steer left or right while rolling. For example Michaud and Caron [20] introduce Roball that is re-prototyped as a toy for child development studies. Seeman et al [4] apply the GroundBot robot to security tasks. Schroll [13] analyzes the dynamics of a single-pendulum spherical robot that also includes conservation of angular momentum enhancement. Zheng et al. [21] analyze the BHQ-1 robot. Park et al [22] proposes a control scheme for a single-pendulum robot. Likewise, Pokhrel et al. [23] introduces SpheRobot, and Hogan et al.i [5] discuss a wind-blown spherical robot that also uses a hanging payload as a single pendulum. Recently, several designs have emerged with two smaller hanging pendulums that allow for stick-slip spin-in-place about the vertical axis. Example designs are proposed and analyzed by Ghanbari et al. [24], Yoon et al. [25], and Wang and Zhao [26]. These single-pendulum designs are more “wheel” than “ball” because they all have a main axis/axle that the robot must roll about. Thus they are not omnidirectional – when facing forwards, they cannot purely roll sideways. However, the single pendulum means that all of the dynamic mass of the robot can be moved very close to the outer shell, maximizing the torque created for a center-of-mass design. A third way of moving the center of mass is by distributed masses sliding along radial, such as tetrahedral, spokes. By sliding the (four) masses as much to one side as possible, the robot rolls in that direction – here, this type of design is called “four-sliders”. Tomik et al. [27] first presented this design along with a geometric motion planner. Javadi and Mojabi [28] present a similar design with a simulation. The main advantage of this design is that it is omnidirectional – the robot can move in any direction instantly. It does not need to steer either via spin or via rolling in a small arc. However, the distributed masses means that some of the dynamic mass is not ideally located for torque generation. For example, if attempting to roll forward, one mass may be located on a spoke aimed backwards and thus not able to apply a positive torque. (Even worse, it may be applying a negative torque.) A proposed but understudied design [29] uses cables and pulleys to move a single mass about the interior of the robot. This design allows for omnidirectional travel and maximized center-of-mass torque, and thus has a lot of potential, but is not heavily explored yet. Here, this design is called the “cable mass” design. Amid and in addition to the mechanism literature, there are several veins of ball-on-plane dynamics applied to spherical robots, motion planning, and spherical robot control research. Besides such work included in the already cited papers – or by similar papers by cited authors – is work done by Der et al. [30], Qiang et al. [31], Peng et al. [32], Tao et al. [33], Ivanova and Pivovarova [34], and Karavaev and Kilin [35]. The
  • 3. 3 Copyright © 2015 by ASME dynamics and control of the four-pendulum mechanism presented here are complicated (as they are for other spherical robots), but they are not discussed in detail in this paper. FOUR PENDULUM DESIGN The spherical robot presented here is a center-of-mass design that uses four pendulum masses in a tetrahedral arrangement. Figure 1 shows a sketch of the (ideal) mechanism, where R is the sphere’s radius. Each pendulum rotates about a tetrahedral direction – changing the angle of the pendulums changes the robot’s center of mass. This mechanism can create rolling torque in any direction instantly, and in practice, the pendulums are rotating in full circles as the robot rolls in any one direction. The main benefit of this design is its omnidirectionality, with larger torque possible than previous tetrahedral designs. The four pendulum design can roll in any direction instantaneously, and can shift its center of mass farther forward than the four-slider design, as will be shown here. Figure 2 shows pictures of a physical prototype. The robot’s electronics and battery are housed in and around central rapid-prototyped tetrahedral box, and the four motors are mounted facing out from that box. The box is centered in the shell using four spokes that reach through gaps between pendulum arcs, and the shell is a 33-cm-diameter plastic ball. Figure 2. Pictures of the prototype four-pendulum spherical robot. The robot uses the Maple r5 as its processor. It is remote controlled using a Futaba 75 MHz RF transmitter and receiver, and carries one battery for the motors (11V 1500mAh LiPo) and a smaller one for the electronics (3.7V 850mAh LiPo). The motors are Pololu 12V brushed DC with a 50:1 reduction and 64 CPR encoders for position feedback. The robot senses orientation via an onboard orientation sensor (CHR-UM6) that processes accelerometer, rate gyro, and magnetic sensor data. Given the desired torque magnitude and direction from the RF data, the robot computes the desired center of mass location in world coordinates, converts that position to robot and then pendulum coordinates, and moves the pendulums to achieve that center. The tetrahedral transformations are as follows (see labels in Fig. 1). The direct angle between axes is the standard tetrahedral angle, φ = cos−1 − 1 3 " # $ % & ' ≈109.5° , (1) and the z-rotation from “xz” pendulum to “+y” pendulum is ϕ =120° . (2) Let the pendulums be identical with length rp rotating on motor axes at tetrahedral distances of rm. Given pendulum i's rotation angle, θi, the position of that pendulum in robot coordinates, xi, is xi = Ri rp cosθ rp sinθ rm ! " # # # # $ % & & & & , (3) where Ri is the rotation matrix for that pendulum. Here, R+z = 1 0 0 0 1 0 0 0 1 ! " # # # $ % & & & , (4) Rxz = cosφ 0 sinφ 0 1 0 −sinφ 0 cosφ " # $ $ $ $ % & ' ' ' ' , (5) R+y = cosϕ −sinϕ 0 sinϕ cosϕ 0 0 0 1 " # $ $ $ $ % & ' ' ' ' • Rxz , (6) R−y = cosϕ sinϕ 0 −sinϕ cosϕ 0 0 0 1 " # $ $ $ $ % & ' ' ' ' • Rxz . (7) For maximum rolling torque, the pendulum length should be as large as possible without collisions between pendulums. That is, larger rp results in larger torque. For this design, the maximum rp without collision is when two pendulum arcs touch – by trigonometry, this occurs when rp = Rsin φ 2 ! " # $ % & ≈ 0.82R rm = Rcos φ 2 ! " # $ % & ≈ 0.58R . (8) Simulations confirm that these dimensions result in the largest possible torque. The following sections compare this mechanism to similar designs with a summary shown in Table 1. The comparisons are made for the ideal, theoretical designs (no static mass, all point-mass dynamic mass). Such a comparison is useful separate from the pragmatic comparison of costs, realistic masses and dimensions, and ease of construction or control.
  • 4. 4 Copyright © 2015 by ASME Design Single Pendulum Four Sliders Cable Mass Four Pendulums Directionality Steered Omni Omni Omni Torque arm, rT <100% <29% <100% <67% Eccentricity, e 100% 0-20% 0 or 100% 36-100% Radius of Gyration, k 100% 0-178% 0-100% 87-216% Table 1. Comparison of Several Theoretical Center-of- Mass Designs DIRECTIONALITY COMPARISON This section compares the four-pendulum design to existing center-of-mass designs in terms of what is called “directionality”. This term is not meant as a technical term and applies to the robot’s shell on the (floor) plane. A sphere on a plane has five degrees-of-freedom: two translational degrees corresponding to the location of the sphere’s center, and three rotational degrees corresponding to the sphere’s orientation at the point. The ideal spherical robot would be able to change any of these degrees instantaneously (and thus be in the technical sense, holonomic). However, a rolling sphere is non-holonomic in that it has no-slip conditions and must roll to change its orientation (and position). An ideal rolling spherical robot would be able to roll in any direction instantaneously without steering – similar to a sports ball rolling on a floor or field. Such omnidirectionality has obvious advantages when navigating complex environments. In this sense, the four-pendulum, sliding- masses, and cable-mass designs are omnidirectional, while the more prevalent single-pendulum designs are not. TORQUE ARM COMPARISON The second comparison to be made between the four- pendulum design and existing center-of-mass designs is in terms of its theoretical normalized torque arm. Here, the torque arm, rT is defined as the effective torque arm of the dynamic center of mass envelope, as a percentage of the sphere radius: rT = horizontal distance to dynamic CoM sphere radius •100% = xCoM 2 + yCoM 2 R •100% . (8) Thus, normalized torque can be 0 to 100% for center-of-mass designs, but larger than that for conservation-of-angular- momentum designs. Each robot implementation has some “static” mass consisting of its components that do not move to change the center of mass and thus the drive torque – its shell, internal framework, and possibly electronics, sensors, battery, and motors. It also has some “dynamic” mass that can be moved to change its center of mass – the pendulums or sliding masses, which may include some or all of the electronics, sensors, battery, and motors. A robot-building design goal is to minimize the static mass while maximizing the dynamic mass, to allow for the maximum possible torque. A theoretical-mechanism design goal is to create as large a torque arm as possible from the dynamic mass, assuming the static mass is negligible. In this sense, the single-pendulum and cable-mass designs have the largest torque arm possible of 100% – if modeled as a point mass, the center of mass can be moved anywhere in the sphere, out to the sphere radius. (In application, the static mass decreases this to a more typical 50- 67% [13].) The four-sliding-mass design, while omnidirectional, has a much smaller torque arm. If the masses are equal, then simulation shows that its center of mass envelope for all possible slider configurations is a polyhedron with a maximum torque arm of 29%. The maximum torque arm also varies based on the direction – in some directions, it is only 20%. The four-pendulum design is also omnidirectional, and has a larger torque arm than the four-slider, but less than the single- pendulum and cable-mass designs. If the masses are equal, then its center of mass can be calculated as the sum of the four xi’s (Eqn. (3)) and the envelope of possible locations looks like a sphere with eight truncations (see Fig. 3). The figure is generated by simulating all possible configurations of the four pendulums, and calculating the resulting dynamic center of mass, but then plotting only the outer envelope. The maximum torque arm in any direction ranges from 56% to 67%. That is, the four-pendulum can create a torque equivalent to moving its dynamic center of mass approximately two-thirds along its sphere radius. Figure 3. Four-pendulum theoretical torque-arm envelope. ECCENTRICITY COMPARISON In addition, this section compares the four-pendulum design to existing center-of-mass designs in terms of the eccentricity, e, of the robot’s rotational moment of inertia, I, about its center. −0.5 0 0.5 −0.5 0 0.5 −0.5 0 0.5 z(normalized) x (normalized) y (normalized)
  • 5. 5 Copyright © 2015 by ASME A general three-dimensional body has a moment of inertia ellipsoid representing its angular inertia in all directions. The axes of the inertia ellipsoid are called the principal axes, and their magnitudes are the principal moments of inertia. The largest inertia, Imajor, about any axis corresponds to the ellipsoid’s semi-major axis, and the smallest, Iminor, corresponds to the ellipsoid’s semi-minor axis. Let Ixx, Iyy, and Izz be the moments of inertia about the x, y, and z axes, and Ixy, Iyz, and Ixz be the corresponding products of inertia. Then, the inertia matrix representing the inertia ellipsoid is I = Ixx −Ixy −Ixz −Ixy Iyy −Iyz −Ixz −Iyz Izz " # $ $ $ $ % & ' ' ' ' . (9) The ellipsoid’s principal axes and values can be found from the standard eigenvector problem Av = λv (10) where generic A is now the inertia matrix I; the eigenvalues, λ, are the magnitude of the principal inertias/axes, and the corresponding eigenvectors, v, are the cosines of the principal axes with respect to the x, y, and z axes. One of the well-known challenges in spherical robots is that the robots tend to wobble as they roll. Angular momentum ( H = Iω ) is conserved, but because the robot’s moment of inertia (I) is non-homogeneous (i.e., an ellipsoid and not a sphere), the angular momentum (ω) also varies. The result is similar to an American football that wobbles outside of a perfect spiral. Thus, having a homogeneous inertia (i.e., a spherical inertia ellipsoid) is advantageous in spherical robots because it reduces wobble. The more non-homogeneous, or “eccentric”, the inertia ellipsoid is, the more the wobble. Here, eccentricity is defined in the standard mathematical sense (also called first eccentricity) but as a percentage: e = 1− Iminor 2 Imajor 2 •100%. (11) In application, any static mass usually decreases eccentricity because the overall mass is more distributed than the pendulum masses. A single-mass (pendulum or cable-controlled) design with a point-mass bob has an inertia ellipse collapsed into a line – its eccentricity is always 100% unless, in the cable-mass design, the mass is in the center of the sphere and the eccentricity has approached 0%. Thus, in any torque situation, these designs have the potential for large wobble if control does not compensate. In terms of eccentricity, the four-slider design is much better. Like the cable-mass design, when all four point masses are at the center of the sphere, the eccentricity is 0%. Otherwise, simulation shows it varies with a maximum of only 20% – the distributed-mass four-slider design is much less prone to wobble than the single-pendulum or cable-mass designs. This effect makes sense – distributing the dynamic mass around the sphere homogenizing the ellipsoid, decreasing the eccentricity. The four-pendulum design is in-between the existing designs. At no point can the four pendulums be moved into the middle like the cable-mass or four-slider design, so the eccentricity is never 0; it’s minimum is 36% (although while still applying a torque). On the other hand, when the four pendulums are paired up on opposite sides of the spheres, the eccentricity is 100% – balanced about the center, but collapsed into a line. A simulation shows that for all pendulum locations, the eccentricity ranges for 36 to 100% with an average of 84%. Thus, the four-pendulum design is more prone to wobble than the four-slider design, but less than the single-mass designs. GYRATION COMPARISON The last comparison to be made between the designs is the radius of gyration – a standard sense of the magnitude of a moment of inertia. The normalized radius of gyration, k, is defined here as percentage of the sphere’s radius: k = I m • 100% R . (12) It can be visualized as the effective radius to a point mass that results in the same moment of inertia. A larger radius of gyration is equivalent to a larger moment of inertia, which can be beneficial when maintaining angular speed, or detrimental when trying to change angular speed. Usually, a design goal is to minimize inertia to allow for a more nimble robot. This is an important difference from minimizing eccentricity. A homogenous sphere has an eccentricity of 0% and a non-zero radius of gyration; the same mass as a point mass can be placed at the same radius of gyration but will now have an eccentricity of 100%. For the single-pendulum design, the maximum theoretical k is always 100% because there is only one point mass, at the sphere’s radius. For the cable-mass design, the maximum k ranges from 0 to 100%, linearly related to the radius to the mass. For the four-slider design, the maximum k varies from 0 (when all masses are pulled in) to 178% (when all masses are extended). For the four-pendulum design, the maximum k varies from 81% to 218% – this design has a large moment of inertia because of the masses cannot collapse into the center. Thus the ranking of the four designs with regards to radius of gyration depends on their configuration, but the cable-mass and four-slider designs have the best possible configurations. PROTOTYPE NUMBERS The physical prototype was designed and built as a proof- of-concept of the four-pendulum driving mechanism, and as an initial testbed for control schemes. As such, it is a success. The robot successfully rolls omnidirectionally as commanded through the RF communication, and maintains (visually) proper heading. It should be noted that the comparisons above are for the theoretical mechanism, and are dampened by the physical application.
  • 6. 6 Copyright © 2015 by ASME For example, the torque arm as calculated (56-67%) is for the ideal mechanism; the physical prototype has a much smaller torque arm. This prototype version is not optimized for dynamic mass over static mass. It has a static mass of shell, batteries, motors, and electronics totaling 2.6kg, with a dynamic mass of 1.6kg – in Chase and Pandya’s [1] terms, its power factor is 0.615. For a 33-cm-diameter sphere, the four pendulums have ideal lengths of around 13.5 cm; because of clearances and non-point-masses, the pendulum have effective lengths of around 10 cm. The effective torque arm of the prototype is closer to 20%. Initial tests with this prototype saw it accelerate with 0.6 m/s2 up to a maximum speed of 0.7 m/s. On the other hand, the eccentricity of the prototype is better than theoretical. The inertia of the static masses means that the inertia ellipse is less dependent on the pendulum locations. The theoretical eccentricity ranges from 36% to 100% with an average of 84%; the prototype’s eccentricity ranges from 30% to 92% with an average of 73%. The radius of gyration also increases with the static mass. For the prototype, it increases to 155-250%. CONCLUSION The four-pendulum omnidirectional spherical mobile robot design has been presented and discussed. Compared to previous center-of-mass designs, the four-pendulum design has large directionality, medium torque arm, medium eccentricity, but large radius of gyration. It is a viable spherical robot design. A physical prototype has been built and tested. It successfully rolls using the pendulums to change the center of mass of the robot. The prototype is successful even with a decreased torque arm and increased radius of gyration, and is less prone to wobble than its theoretical equivalent. Future work includes rigorous dynamical analysis, robust control implementations, and a more optimized prototype. The comparisons also suggest that the cable-mass mechanism is a beneficial design and should be analyzed further. ACKNOWLEDGMENTS This work was funded in part by the Central Michigan University FRCE Grant #48868. REFERENCES [1] Chase, R., and Pandya, A., 2012. “A review of active mechanical driving principles of spherical robots”. Robotics, 1, pp. 3–23. [2] Crossley, V. A., 2006. A literature review of the design of spherical rolling robots. [3] Armour, R. H., and Vincent, J. F. V., 2006. “Rolling in nature and robotics: A review”. Journal of Bionic Engineering, 3(4), 12, pp. 195–208. [4] Seeman, M., Broxvall, M., Saffiotti, A., and Wide, P., 2006. “An autonomous spherical robot for security tasks”. In IEEE Conference on Computational Intelligence for Homeland Security and Personal Safety. [5] Hogan, F. R., Forbes, J. R., and Barfoot, T. D., 2014. “Rolling stability of a power-generating tumbleweed rover”. Journal of Spacecraft and Rockets, 51(6), pp. 1895–1906. [6] Hernandez, J. D., Barrientos, J., del Cerro, J., Barrientos, A., and Sanz, D., 2013. “Moisture measurement in crops using spherical robots”. Industrial Robot: An International Journal, 40(1), pp. 59–66. [7] Li, M., Guo, S., Hirata, H., and Ishihara, H., 2015. “Design and performance evaluation of an amphibious spherical robot”. Robotics and Autonomous Systems, 64, pp. 21–34. [8] Michaud, F., Laplante, J.-F., Larouche, H., Duquette, A., Caron, S., Letourneau, D., and Masson, P., 2005. “Autonomous spherical mobile robot for child-development studies”. IEEE Transactions on Systems, Man, and Cybernetics, 35(4), July, pp. 471–9. [9] Brown, H. B. J., and Xu, Y., 1997. “A single-wheel gyroscopically stabiliszed robot”. IEEE Robotics and Automation Magazine, September. [10] Bhattacharya, S., and Agrawal, S. K., 2000. “Spherical rolling robot: A design and motion planning studies”. IEEE Transactions of Robotics and Automation, 16(6), pp. 835–9. [11] Shu, G., Zhan, Q., and Cai, Y., 2009. “Motion control of spherical robot based on conservation of angular momentum”. In ICMA International Conference on Mechatronics and Automation, pp. 599–604. [12] Joshi, V. A., Banavar, R. N., and Hippalgaonkar, R., 2010. “Design and analysis of a spherical mobile robot”. Mechanism and Machine Theory, 45(2), 2, pp. 130–136. [13] Schroll, G. C., 2010. “Dynamic model of a spherical robot from first principles”. Master’s thesis, Colorado State University, Summer. [14] Artusi, M., Potz, M., Aristizabal, J., Menon, C., Cocuzza, S., and Debei, S., 2011. “Electroactive elastomeric actuators for the implementation of a deformable spherical rover”. IEEE/ASME Transactions on Mechatronics, 16(1), pp. 50–57. [15] Wait, K. W., Jackson, P. J., and Smoot, L. S., 2010. “Self locomotion of a spherical rolling robot using a novel deformable pneumatic method”. In IEEE International Conference on Robotics and Automation. [16] Sugiyama, Y., Shiotsu, A., Yamanaka, M., and Hirai, S., 2005. “Circular/spherical robots for crawling and jumping”. In IEEE International Conference on Robotics and Automation. [17] Halme, A., Torsten, and Wang, Y., 1996. “Motion control of a spherical mobile robot”. In 4th International Workshop on Advanced Motion Control, pp. 259–64. [18] Bicchi, A., Balluchi, A., Prattichizzo, D., and Gorelli, A., 1997. “Introducing the Sphericle: an experimental testbed for research and teaching in nonholonomy”. In IEEE International Conference on Robotics and Automation, pp. 2620–5. [19] Alves, J., and Dias, J., 2003. “Design and control of a spherical mobile robot”. Journal of Systems and Control Engineering, 217, September, pp. 457–467. [20] Michaud, F., and Caron, S., 2002. “Roball, the rolling robot”. Autonomous Robots, 12(2), pp. 211–222. [21] Zheng, M., Qiang, Z., Jinkun, L., and Yao, C., 2011. “Control of a spherical robot: Path following based on
  • 7. 7 Copyright © 2015 by ASME nonholonomic kinematics and dynamics”. Chinese Journal of Aeronautics, 24(3), pp. 337–345. [22] Gye-Do Park, K.-H. K., Ba Da, Y., and Lee, J.-M., 2011. “Roll control of spherical robot using the IMU”. In 3rd SPENALO International Symposium. [23] Pokhrel, D., Luitel, N. R., Das, S., and Ray, D. N., 2013. “Design and development of a spherical robot (SpheRobot)”. In 1st International and 16th National Conference on Machines and Mechanisms. [24] Ghanbari, A., Mahboubi, S., and Fakhrabadi, M., 2010. “Design, dynamic modeling and simulation of a spherical mobile robot with a novel motion mechanism”. In Mechatronics and Embedded Systems and Applications (MESA), IEEE/ASME International Conference on, pp. 434– 439. [25] Yoon, J.-C., Ahn, S.-S., and Lee, Y.-J., 2011. “Spherical robot with new type of two-pendulum driving mechanism”. In Intelligent Engineering Systems (INES), 15th IEEE International Conference on, IEEE, pp. 275–279. [26] Wang, L., and Zhao, B., 2015. “Dynamic modeling and control strategy for turning in place motion of a two-coaxial pendulums driven spherical inspector based on stick–slip principle”. Mechanism and Machine Theory, 83, pp. 69–80. [27] Tomik, F., Nudehi, S., Flynn, L. L., and Mukherjee, R., 2012. “Design, fabrication and control of SpheRobot: A spherical mobile robot”. Journal of Intelligent Robot Systems, 67(2), pp. 117–131. [28] Javadi A., A. H., and Mojabi, P., 2004. “Introducing glory: A novel strategy for an omnidirectional spherical rolling robot”. Journal of Dynamic Systems, Measurement, and Control, 126(3), pp. 678–83. [29] Lux, J., 2005. Alternative way of shifting mass to move a spherical robot. Tech. rep., NASA’s Jet Propulsion Laboratory, June. [30] Der, R., Martius, G., and Hesse, F., 2006. “Let it roll — emerging sensorimotor coordination in a spherical robot”. In International Conference on the Simulation and Synthesis of Living Systems, MIT Press, pp. 192–198. [31] Qiang, Z., Zengbo, L., and Yao, C., 2008. “A back- stepping based trajectory tracking controller for a non-chained nonholonomic spherical robot”. Chinese Journal of Aeronautics, 21(5), pp. 472–480. [32] Peng, Y.-F., Chiu, C.-H., Tsai, W.-R., and Chou, M.-H., 2009. “Design of an omni-directional spherical robot: Using fuzzy control”. In The International Multiconference of Engineers and Computer Scientists, Vol. 1, pp. 18–20. [33] Tao, Y., Hanxu, S., and Qingxuan, J., 2012. “Variable structure control of pendulum-driven spherical mobile robots”. In 3rd International Conference on Computer and Electrical Engineering (ICCEE 2010 no. 2). [34] Ivanova, T., and Pivovarova, E., 2013. “Dynamics and control of a spherical robot with an axisymmetric pendulum actuator”. Nonlin. Dyn. Mob. Robot, 1(1), pp. 71–85. [35] Karavaev, Y. L., and Kilin, A. A., 2015. “The dynamics and control of a spherical robot with an internal omniwheel platform”. Regular and Chaotic Dynamics, 20(2), pp. 134–152.