The 2010 IEEE/RSJ International Conference on

Intelligent Robots and Systems

October 18-22, 2010, Taipei, Taiwan

Dynamics and motion control of a two pendulums driven spherical

robot
Bo Zhao, Mantian Li, Haitao Yu, Haiyan Hu, Lining Sun

Abstract—This paper deals with the dynamics and motion protected by the outer shell [2]. So spherical robot can be
control of a spherical robot designed for reconnaissance and used in hostile environments, such as rescue in disaster and
unstructured hostile environment exploration. The robot in this military reconnaissance, especially planetary exploration.
paper has three DOFs and two inputs, of which the nature is a According to the differences of actuator, the representative
nonlinear and underactuated system with nonholonomic spherical robots are grouped into six types: rotor type, car
dynamic constraints. The improved construction of two type, mobile masses type, gyroscope type, memory alloy
pendulums offers novel motion principle of spherical robot,
type and pendulum type; each type has merits and demerits
which is moving simultaneously actuated by both eccentric
moment and inertial moment generated by the two pendulums. respectively [3]. The basic motion principle of the six types
Meanwhile the mobility is enhanced when the robot behaves can be summarized as two kinds of motion principle; one is
dynamically. The emphasis is placed on the linear motion and generating friction force by driving the device which is
turning in place motion control. The dynamic model of linear contact with the inner surface of the shell, such as rotor type
motion is formulated on the basis of Lagrange equation, and a and car type; the other is continuously changing the position
smooth trajectory planning method is proposed for linear of center of gravity of the robot by controlling the drive unit,
motion. A feedback controller is constructed to ensure the such as mobile masses type, gyroscope type, memory alloy
accurate trajectory planning. Turning in place motion is an type and pendulum type. The former is applied in the early
indispensable element of omni directional locomotion which
spherical robots; the latter is widely utilized by recent
can enhance the mobility of spherical robots. The dynamic
model of turning in place motion is derived on the theory of spherical robots, especially the pendulum type which is
moment of momentum, and a stick-slip principle is analyzed. employed by most spherical robots. The “roball” spherical
The two motion control methods are validated by both robot developed by Francois Michaud, the “BHQ”
simulations and prototype experiments. developed by ZhanQiang and the “GroundBot” developed
by Mattias Seeman are all based on pendulum type [4]–[6].
I. INTRODUCTION The pendulum which is driven by two motors can rotate

T raditionally, mobile robot is considered to be wheeled about the horizontal axis and the vertical axis, so the
robot or legged robot. The wheeled robot features great eccentric moment and inertia moment are generated, by
mobility on relatively flat terrains, and the legged robot which the robot can roll linearly or steer. The drawback of
has potentially greater mobility on tough terrain [1]. In pendulum driven spherical robot is that the energy input of
recent years, spherical robot, a new type of mobile robot, has the motor for vertical rotation of pendulum is utilized only
attracted wide attention of researchers. Compared with for steering, which is not an effective utilization of power
wheeled or legged robots, besides great mobility on various supply. The researches on spherical robot in recent years
terrains, spherical robot has the following advantages: (1) mainly focus on trajectory planning and motion control
moving without roll-over problem, the stability can be [7]–[10], rather than motion principle and mechanism design.
restored by itself, the attitude is easy to recover after However, the linear motion and turning in place motion have
collision or falling down from the height; (2) the spherical not been proposed as the individual motion in previous
robots can potentially move omnidirectionally with high researches on trajectory planning and motion control.
d8exterity, such as turning in place; (3) entire system is This paper deals with the dynamics and motion control of
accommodated inside the ball-shaped shell, all devices are a spherical robot based on novel improved two pendulums
type which engages novel motion principle. The drive
mechanism of the robot in this paper differs from the
Manuscript received Mar. 10, 2010. This work was supported in part by previous designs. The linear motion control and turning in
High Technology Research Development Program of China place motion control are studied respectively as two separate
(No.2005AA4202302) and the Program for Changjiang Scholars and motions. The design details of the robot are described in Sec.
Innovative Research Team in University (IRT0423).
Bo Zhao is with the State Key Laboratory of Robotics and System, II along with the motion principle. Section III provides
Harbin Institute of Technology. Harbin, 150001 China (corresponding linear motion control of the robot. The dynamic equation of
author to provide phone: 13674681656; fax: 0451-86402217; e-mail: the linear motion is formulated on constrained Lagrange
[email protected]). method and principle of virtual work. Based on the
Mantian Li is with the State Key Laboratory of Robotics and System,
Harbin Institute of Technology. Harbin, 150001 China (e-mail: dynamics, a smooth trajectory planning control method is
[email protected]). proposed. A feedback controller is constructed for the
Haitao Yu is with the State Key Laboratory of Robotics and System, accurate trajectory planning. The effectiveness of the
Harbin Institute of Technology. Harbin, 150001 China (e-mail: method is validated by simulation and prototype experiment.
[email protected]).
Section IV provides the turning in place motion control. The

978-1-4244-6676-4/10/$25.00 ©2010 IEEE 147

Authorized licensed use limited to: UNIVERSIDAD AUTONOMA METROPOLITANA. Downloaded on November 08,2024 at 23:16:27 UTC from IEEE Xplore. Restrictions apply.
 dynamic equation is derived on the theory of moment of angle θ, which results in the displacement of the center of
momentum. A stick-slip principle is interpreted in detail. gravity of the robot. The robot could only roll about the x
The control method is verified by simulation, and the results axis under the action of the eccentric moment. Meanwhile
of prototype experiment validate the effectiveness of the the motor should keep rotating to maintain the angle θ for
control method. the continuous rotation about x axis, and the inertia forces
are simultaneously generated by the acceleration of the
II. SYSTEM DESCRIPTION pendulums, which are paralleled to yoz plane. The inertia
forces could be decomposed into two orthogonal forces
A. Mechanical Structure
along y and z axes and one couple about x axis. Projected
The mechanical structure of the robot is shown in Fig. 1, onto the yoz plane, the inertia forces along the y and z axes
which is mainly composed of two motors, two pendulums, can afford inertia moment about the z and y axes
springs, linear bearings, two guides and the outer shell. respectively, so the robot could roll about the z and y axes.
The inertia moment about the z axis is sufficient to activate
the robot turning in place on relatively smooth ground.
Because of static equilibrium of the moment about y axis
and the restriction from the shape of outer shell, the robot
can only roll about the y axis in a certain angle. The energy
input is predominantly utilized for reacting against the
friction force in slow or stable motion, and the inertia
moment is significant with respect to eccentric moment in
dynamically unstable motion.
C. Control system

Fig. 1. Mechanical structure of the robot

The outer shell is ellipsoid, whose diameter of major axis

is 176 mm, and the minor axis 146 mm. While in irregular
motion, such as passing over bumps, the attitude of the robot
is easy to be observed. Meanwhile, the pitch angle is
restricted to a certain range by the ellipsoid shell, which
ensures the robot can tolerate side slopes without rolling
over. Fig. 2. Control system diagram of the robot
A novel two pendulums type is adopted. The robot is
driven by two pendulums placed diametrically opposite on The two classes control system is designed because the
the major axis of the ellipsoid shell, which are actuated by robot belongs to a nonholonomic and under-actuated system.
two direct current (DC) servomotors. The two pendulums Fig. 2 is the control system diagram of the robot, which is
can be only rotated about the major axis, which could composed of sensors, communication module, control
change the position of the center of gravity of the robot and module and actuators. The module of sensor consists of
meanwhile afford the inertia force. This configuration brings encoders of the servomotors, angular rate sensors, angular
a novel motion principle of spherical robot. The robot is acceleration sensors and directional gyro, which are all
motivated by eccentric moment, inertia force and inertia micro electromechanical systems (MEMS) devices. An
moment simultaneously. The total weight of the masses is embedded control system is designed for attitude computing
700 gram which accounts for 42.9% of the total mass of and motion planning. The operator can not directly control
robot. Compared with the traditional pendulum type, the the pendulums to activate the robot, because the
improved type can afford more eccentric moment and inertia point-contact between outer shell and ground and small error
force which make the robot has featured high speed, the in motions of the two pendulums may cause the unexpected
ability of slope climbing and obstacle overcoming. The drive motion of the robot. So the instructions for motion of robot
mechanism of the robot is composed of two motors and two sent by a remote controller should be “move straight” or
pendulums which are connected to the linear guide by linear “turn right for 20 degree”. The data of encoders, angular rate
bearings whose two ends are pressed by springs. This elastic sensor and directional gyro are processed by attitude
connection can buffer the impact when the robot falls down computing algorithms to get the current attitude of robot.
from the high or is in collision. And in the future works, the The instructions for motors are computed by motion
masses will be controlled to resonate with the springs in planning algorithms and then sent to actuators.
order to let the robot jump while in trap.
III. LINEAR MOTION CONTROL
B. Motion Principle
A. Dynamic model of linear motion
As shown in Fig. 1, the spindle is along the x axis. The
two pendulums can only rotate about the spindle to a tilt In linear motion, the robot can be projected onto side view,
which is modeled as a pendulum inside a rigid sphere

148
Authorized licensed use limited to: UNIVERSIDAD AUTONOMA METROPOLITANA. Downloaded on November 08,2024 at 23:16:27 UTC from IEEE Xplore. Restrictions apply.
 unrelated to the ellipsoid shell as shown in Fig. 3. The d ∂L ∂L
dynamics of the robot is derived under the following ( )− = Qj (5)
dt ∂q j ∂q j
assumptions: (i) there is no slip between the shell and floor.
(ii) the two pendulums rotate synchronously without angle The dynamics of the linear motion can be written as
difference. ⎧ τ
y + mrθ cos θ − mrθ 2 sin θ =
⎪(m + 2M ) 
⎨ R (6)
⎪ myr
 cos θ + mr 2 
θ + mgr sin θ = τ
⎩
This dynamic function is not integrable, but provides the
basis for trajectory planning that can make the system
controllable from an initial configuration to a desired
configuration in practical application.
B. Smooth trajectory planning of linear motion
The smooth trajectory planning from an initial
configuration to an expected final configuration is essential
to the robot, which can enhance the performance of linear
motion. The dynamic function given by (6) can be
rearranged as
Fig. 3. Planar simplified model of linear motion
y = f (θ ,θ,θ)

As shown in Fig. 3, the coordinate y0oz0 is considered to
be reference coordinates fixed to the ground, and the mgr sin θ + mr sin θ Rθ 2 + (mr 2 − mr cos θ R)θ (7)
=
coordinate yoz is fixed to the robot. The two pendulums are (m + 2M ) R − mr cos θ
rotated by motors to a θ angle to change the position of The coefficients of the terms related to θ in (7) are
center of the sphere. Additionally the inertia forces in yoz constant terms. This indicates that in linear motion, in order
plane perpendicular to the pendulums are generated. The to keep a desired uniform speed (  y = 0 ) of the robot, the
sphere rolls about the x axis driven by the eccentric moment pendulums should be maintained on the vertical position
and inertia moment from the two pendulums, so the center of
the sphere moves straightly along the y0 axis. The ( θ = 0 , θ = 0 , sin θ = 0 ). The nonzero tilt angle of the
locomotion should be maintained by the continuous rotation pendulums may activate the robot to accelerate. It is also
of the two pendulums. The Lagrange equations are used to implied that the desired velocity of the robot in linear motion
derive the dynamic equations of linear motion. The angle can be obtained by given the proper tilt angle, that is, the
between the pendulums and vertical is referred as the tilt position of the robot is controllable.
angle θ. The angle the sphere rotates through with respect to The whole process of linear motion from initial static
the reference coordinates is the body angle α. So the angle configuration to final static configuration is divided to three
the motor shafts rotates through is θ+α. The tilt angle θ and stages, which are starting stage, uniform motion stage and
distance of the center of the sphere y are considered to be the stopping stage. The corresponding curve of velocity y
generalized coordinates in Lagrange equations, and the should fit to the ladder diagram. From the starting stage to
torque of the motors τ is the generalized force. uniform motion stage, the velocity y accelerates from zero to
The kinetic energy of the sphere except for pendulums a desired uniform velocity y d , so the pendulums should be
respect to ground can be expressed as rotated to the desired angle θd first, then lean back to the
1 1 vertical position (Fig. 4(a)). With the realization of these
T1 = My 2 + J α 2 (1)
2 2 features of the system of linear motion, inspired by [11] and
where M is the mass of the sphere, J = MR 2 is the moment [12], a smooth trajectory for tilt angle of the pendulums
of inertia of the sphere, R is the sphere radius, y = α R is based on normal distribution function is proposed by (8):
2 2
the forward speed of the center of the sphere. θ (t ) = θ1e − μ ( t −t
+ θ 2 e − μ 2 ( t − ts 2 )
1 s1 )
(8)
The kinetic energy of the pendulums can be expressed as where θ1 and θ2 are the amplitudes of the normal distribution
1 1 function, ts1 and ts2 are the time the tilt angle reaches the
T2 = m( y + rθ sin θ )2 + m(rθ cos θ ) 2 (2)
2 2 peak, μ1 and μ2 are the constant scalars. The constant scalars
where m is the mass of the two pendulums, r is the radius of μ1 and μ2 determine the width of the normal distribution
the pendulums. function. Regulation of these parameters can change the
With respect to the center of the ball considered to be the state of the robot in linear motion.
point of zero potential, the potential energy of the system is Numerically solved by MATLAB, the curve of tilt angle
V = − mgr cos θ (3) of pendulums is shown in Fig. 4(a). The robot is controlled
The Lagrange function can be express as from the initial static configuration to the final static
configuration through the smooth trajectory. Fig. 4(b) is the
L = T1 + T2 − V (4)
velocity and displacement of the sphere under the action of
According to Lagrange equations the pendulums. The accelerating state, uniform speed motion
stage and decelerating stage are described clearly.

149
Authorized licensed use limited to: UNIVERSIDAD AUTONOMA METROPOLITANA. Downloaded on November 08,2024 at 23:16:27 UTC from IEEE Xplore. Restrictions apply.
 the simulation time is 20s, the planned tilt angle θ tracks the
desired angle θd. as shown in Fig. 6. The solid curve is the
tilt angle in simulation and the dashed line is the desired
angle.
The effect of the feedback controller can be found in Fig.
7. The displacement and velocity of the robot output from
the controller in simulation are shown in Fig. 7(a) and Fig.
7(b) respectively. The feedback controller is capable to
control the robot to move linearly in the expected motion
(a) Tilt angle of pendulums (b) Velocity and displacement of sphere
Fig. 4. Smooth trajectory planning for linear motion under the disturbance from the floor.

C. Feedback controller for linear motion D. Prototype experiment of linear motion

Fig. 5. Feedback controller for linear motion

The smooth trajectory planning method is theoretically

proposed. Actually, the unexpected noise from floor
condition, uncertain friction between sphere and floor and
mechanical error may disturb the state of the robot. A
feedback controller is designed to ensure the accurate Fig. 8. Linear motion experiment
trajectory planning for the system. The diagram of the The linear motion experiment is done in the indoor
controller is shown in Fig. 5. The desired tilt angle θd is environment in order to minimize the disturbance from the
regulated by function (8). The controller feeds back the floor. The smooth trajectory planning was implemented on
practical tilt angle θ and outputs the torque τ of the motor the prototype. The two pendulums are controlled to rotate
shafts actuated on the pendulums. A nonlinear friction synchronically to ensure the linear route. As shown in Fig. 8,
compensation model is added to adjust the torque. The the robot moves from the initial static configuration to the
friction model is composed of viscous damping friction term final static configuration by the tilt of the pendulums. The
and impact force term which are related to the velocity of the tilt angle of the pendulums is calculated by integrating the
sphere and displacement of the sphere respectively. angular velocity of the sphere and data measured by
encoders of the motors. The curve of tilt angle is plotted in
Fig. 9. It can be concluded that the practical curve of tilt
angle coincides with the smooth trajectory planning in
theory.

Fig. 6. Tilt angle

Fig. 9. Tilt angle in experiment

The displacement and velocity of the robot in experiment

are shown in Fig. 10. The velocity of the sphere in Fig. 10(b)
is measured by the angular rate sensor, and the data of
(a) Displacement of the sphere (b) Velocity of the sphere
Fig. 7. Angle of the sphere in simulation displacement is obtained by integration of the data of
velocity. The noise on the curves may come from the
The controller is verified by simulation. Given the value sensitivity of the angular rate sensor and the condition of the
of the parameters, θ1 =1, θ2 =-1, μ1 =μ2 =0.8, ts1 =4, ts2 =16,

150
Authorized licensed use limited to: UNIVERSIDAD AUTONOMA METROPOLITANA. Downloaded on November 08,2024 at 23:16:27 UTC from IEEE Xplore. Restrictions apply.
 floor. It can be concluded that the motion state of the sphere about the z axis in horizontal plane xoy. So the velocity of
coincides with the smooth trajectory planning. one pendulum can be written as
⎧⎪Vr = rθ
⎨ (10)
⎪⎩Vd = r ′ϕ
where Vr is the linear velocity of pendulum about x axis,
Vd is the linear velocity about z axis, θ is the rotation angle
of the pendulums about the rod, φ is the rotation angle of the
robot about z axis, and r ′ = d 2 + r 2 sin θ is the rotation
radius about z axis.
(a) Displacement of the sphere (b) Velocity of the sphere The projection of the two velocities in (10) in the
Fig. 10. Angle of the sphere in experiment orthogonal plane can be expressed as
V⊥ = rθ cos θ cos α − r ′ϕ (11)
IV. TURNING IN PLACE CONTROL where α is the acute angle between the rod and the line
Some spherical robots have featured turning in place connecting the two cancroids of the pendulums on top view.
motion, which can enhance the mobility and tolerance of The moment of momentum of the system is given by
hostile environment. This motion exists in the spherical d (12)
LZ = 2mr ′(rθ cos θ cos α − ϕ ) − J ϕ
robots which belong to statically stable system, such as rotor cos α
type, car type and pendulum type. The turning in place According to the theorem of moment of momentum, the
motion has not been proposed separately as a definite motion equivalent friction moment can be calculated by
in previous study, which should not be neglected as one of d⎛  d ⎞ (13)
motion characteristics of spherical robots.
Mf = ⎜ 2mr ′(rθ cos θ cos α − ϕ ) − J ϕ ⎟
dt ⎝ cos α ⎠
A. Dynamics of turning in place motion The dynamics of turning in place motion is given by
t
d2 (14)
2mdrθ cos θ − 2m ϕ − J ϕ = ∫ M f dt
cos 2 α 0

where d2 d2 .
=
cos α cos 2 (tan −1 ( r sin θ ))
2

d
The coupling trigonometric item is simplified by
⎛ r sin θ ⎞
cos 2 ⎜ tan −1 ( ) ⎟ ≈ k cos 2θ + b (15)
Fig. 11. Model of turning in place ⎝ d ⎠
The model of turning in place motion is shown in Fig. 11. The fitting coefficient k and b are determined by the
The major axis is simplified as a rod fixed to the sphere. The structure parameters r and d.
two pendulums can only rotate about the x axis on the So the dynamic function can be simplified as
t
endpoints of the rod O1 and O2. When the two pendulums 2md 2 (16)
2mdrθ cos θ − ϕ − J ϕ = ∫ M f dt
rotate synchronously to opposite direction, the eccentric (k cos 2θ + b) 0

moment about the x and y axes and the inertia force along The relationship between φ, θ and Mf is given by
the z axis are in equilibrium. The force condition of turning P+Q
sin θ (t ) +
in place motion is given by mdr 2Q
t
M f tdt (17)
ϕ= Ln −∫
∑ Fx = 0 P( P + Q)
sin θ (t ) −
P+Q 0 J + 2 md 2 ( k cos 2θ + b )

∑ Fy = Fgy1 − Fgy1 = 0 2Q

∑ Fz = Fgz1 + Fgz1 − G + N = 0 (9)

where P and Q are the constant terms once the k and b are
determined, P=J+2bmd2, Q=2kmd2.
∑ M x = M gx1 − M gx2 = 0 The dynamic equation (17) can not be solved by analytical
∑ M y = M gy1 − M gy2 = 0 method, but is essential for verifying the effectiveness of the
∑ M z = M gz − M f stick-slip control method.
It can be concluded that the sphere is actuated only by the B. Stick-slip control of turning in place motion
inertia moment about z axis generated by the inertia forces
along y axis, so the sphere can only roll about z axis in
theory, which is turning in place.
The motions of the sphere and two pendulums can not be
projected to one plane, so the dynamics of turning in place
motion is derived on the theorem of moment of momentum.
The motion of one pendulum can be decomposed to one
rotation about the rod in vertical plane yoz and one rotation
Fig. 12. Stick stage model

151
Authorized licensed use limited to: UNIVERSIDAD AUTONOMA METROPOLITANA. Downloaded on November 08,2024 at 23:16:27 UTC from IEEE Xplore. Restrictions apply.
 The stick-slip control method is proposed for turning in ⎧ θm 1
place motion, which can be divided into stick stage and slip ⎪ − K cos(ωt ) + (1 − K )θ m stick
⎪ v v (19)
stage. In initial state, the sphere is static, and the pendulums θ (t ) = ⎨
θ
⎪ m cos( K ωt ) + (1 − 1 )θ
are in vertical position with no power supplement. Fig. 12 is slip
⎪⎩ K v f
Kv
m
the stick stage model of turning in place motion. The two
pendulums are rotated in a same angular velocity but where θm is the maximum value of tilt angle, ω is the
opposite direction. The inertia moment generated by inertial harmonic frequency, Kv is the angular velocity coefficient
force of the acceleration of pendulums at the time t0+ is which dominates the angular velocity of pendulums, Kf is
insufficient to activate the sphere, that is, the sphere is kept the frequency coefficient which can change the time
static by the static friction moment. The angle θ increases to proportion of slip stage in one step to regulate the value of
a certain value slowly with time t until the time t1. inertial force. The value of parameters in (19) are given,
θ m = π / 3 ， ω = π / 0.8 ，Kv=2，Kf=4.
Fig. 14 is the results of the simulation. The solid line and
dashed line are experimental results and desired values
respectively. The time of one step of stick-slip process is 1s.
In simulation, the noise from uncertain factors is added to
the model. The equivalent friction moment is sufficient in
simulation to keep the robot static in stick stage. The tilt
angle displacement of one pendulum is shown in Fig. 14(a).
Fig. 13. Slip stage model
The two pendulums are controlled to rotate periodically to
The slip stage is shown in Fig. 13. The pendulums are an opposite direction. The slowly rising curve of tilt angle is
rotated to angle θ of opposite direction at the time t1. In slip the stick stage, the robot is static, and so the corresponding
stage, the pendulums are controlled back to the vertical body turning angle is zero. The suddenly falling curve is the
position rapidly in a short interval. The sphere is actuated to slip stage, the robot turns about the vertical axis to a certain
turn in place by the inertia moment generated by inertia angle by the inertia moment generated by the acceleration of
force of large acceleration of pendulums, that is, the friction the pendulums as shown in Fig. 14(b).
moment is weak compared with inertia moment in slip stage.
When the pendulums return back to vertical position at the
time t2, the power supplement to motors is off. Then the
sphere will stop when the kinetic energy is exhausted by the
friction moment. The two stages constitute one step of the
turning in place motion.
C. Simulation of stick-slip control
Considered the complexity of the dynamics of turning in (a) Tilt angle displacement (b) Displacement of sphere
place motion, relationship between the parameters is solved Fig. 14. Simulation of turning in place motion
by simulation. The system parameters of robot in simulation D. Prototype experiment of turning in place motion
given in TABLE I are experimentally determined.
TABLE I
PARAMETERS OF TURNING IN PLACE IN SIMULATION

Symbol Quantity Value

M mass of the sphere 0.95 kg

mass of one
m 0.35 kg
pendulum
moment of inertia of
J 0.0029 kgm2
the sphere
r radius of pendulums 0.038 m
d half length of the rod 0.045 m
k fitting coefficient -0.3082
b fitting coefficient 1.3082

The friction model in simulation is given by

⎧ D ϕ sign(ϕ ) , ϕ ≠ 0 Fig. 15. Experiment of turning in place motion
Mf =⎨ v (18)
⎩ M static , ϕ = 0
The stick-slip principle is verified in experiment of
where Mstatic is the static friction moment, Dv is the viscous turning in place motion. The two pendulums are controlled
damping coefficient. The value is tested by experiment on to the opposite direction. The robot can only rotate about
flat indoor floor, Mstatic=0.01424Nm, Dv=0.057Nms/rad. vertical axis without rolling in theory, so the tilt angle is the
The tilt angle θ is controlled by angle the motor shafts rotate through. As shown in Fig. 15,

152
Authorized licensed use limited to: UNIVERSIDAD AUTONOMA METROPOLITANA. Downloaded on November 08,2024 at 23:16:27 UTC from IEEE Xplore. Restrictions apply.
 the robot can continuously turn in place by the repeated REFERENCES
input of the two stages of stick-slip principle. The robot can [1] Y. S. Xu and S. K. Au, “Stabilization and path following of a single
orient to the desired direction through the regulation of the wheel robot,” IEEE trans. Mechatronics, vol. 9, no.2, pp. 407–419,
amplitude of the tilt angle of the two pendulums. Jun. 2004.
[2] A. Halme, T. Schonberg, Y. Wang, “Motion control of a spherical
mobile robot.,” IEEE International Workshop on Advanced Motion
Control, AMC, vol.1, pp.259–264, 1996.
[3] R. H. Armour and J. F. V. Vincent, “Rolling in Nature and Robotics:
A Review,” Journal of Bionic Engineering, vol.3, no.4, pp.195−208,
Dec. 2006
[4] F. Michaud and S. Caron, “Roball, the rolling robot,” Autonomous
Robots, vol. 12, no. 2, pp. 211–222, Mar. 2002.
[5] Q. Zhan, C. Jia, X. H. Ma and Y. T. Zhai, “Mechanism Design and
Motion Analysis of A Spherical Mobile Robot,” Chinese Journal of
Mechanical Engineering, vol.18, no. 4, pp.542–545, Dec. 2005.
[6] M. Seeman, M. Broxvall, A. Saffiotti and P. Wide, “An autonomous
spherical robot for security tasks,” Proceedings of the 2006 IEEE Int.
Conf. Computational Intelligence for Homeland Security and
Personal Safety, pp. 51–55, Oct. 2006.
[7] S. Bhattacharya and S. K. Agrawal, “Spherical Rolling Robot: A
Design and Motion Planning Studies,” IEEE trans. Robotics and
Automation, vol. 16, pp.835–839, 2000.
[8] R. Mukherjee, M. A. Minor and J. T. Pukrushpan, “Motion Planning
for a Spherical Mobile Robot: Revisiting the Classical Ball-Plate
Fig. 16. Angular velocity of the robot Problem,” Journal of Dynamic Systems, Measurement and Control,
Transactions of the ASME, vol. 124, pp. 502–511, Dec. 2002.
The angular velocity of the robot about vertical axis is
[9] T. Das and R. Mukherjee, “Reconfiguration of a Rolling Sphere: A
shown in Fig. 16. The curve is like sawtooth wave. In the Problem in Evolute-Involute Geometry,” Journal of Applied
first stick stage, the angular velocity is negative because in Mechanics, Transactions ASME, vol. 73, no. 4, pp. 590–597, Jul.
initial stage, the instantaneous momentum conservation 2006.
[10] T. Otani, T. Urakubo, S. Maekawa and H. Tamaki, “Position and
occurs between the sphere and the pendulums, the turning
Attitude Control of a Spherical Rolling Robot Equipped with a Gyro,”
angle and the inertia moment of pendulums are to the IEEE International Workshop on Advanced Motion Control, AMC, pp.
opposite direction. It can be found that in experiment, the 416–421, 2006.
turning angle of the sphere is not zero in stick stage. The [11] U. Nagarajan, A. Mampetta, G. A. Kantor and R. L. Hollis, “State
Transition, Balancing, Station Keeping, and Yaw Control for a
reason is that the friction moment is insufficient to stop the
Dynamically Stable Single Spherical Wheel Mobile Robot,” 2009
robot immediately. IEEE Int. Conf. Robotics and Automation, Kobe, May. 2009, pp.
998–2003.
V. CONCLUSION [12] U. Nagarajan, G. Kantor and R. L. Hollis, “Trajectory Planning and
Control of an Underactuated Dynamically Stable Single Spherical
A spherical robot of new drive mechanism is studied in Wheeled Mobile Robot,” 2009 IEEE Int. Conf. Robotics and
this paper. The linear motion and turning in place motion Automation, Kobe, May. 2009, pp. 3743–3748.
control of a two pendulums driven spherical robot are
studied in this paper. The emphasis of linear motion control
is placed on the movement to a desired configuration. The
relationship between dynamic parameters is found based on
the dynamic equation. The smooth trajectory planning
method for linear motion based on normal distribution
function is proposed. A feedback controller is constructed
for the accurate trajectory planning. The tilt angle and sphere
displacement in simulation and experiment represent
accordance with the expectation. The robot can be controlled
to move from an initial static configuration to a final static
configuration through the adjustment of parameter. To
enhance the mobility of the robot, a stick-slip control
method is proposed for turning in place motion. The
dynamics deduced on the theory of moment of momentum
provides the basis for validating the principle. The two
stages of one step are interpreted in detail. The principle is
verified by simulation. In prototype experiment, the robot
can continuously turn in place to a desired angle.

153
Authorized licensed use limited to: UNIVERSIDAD AUTONOMA METROPOLITANA. Downloaded on November 08,2024 at 23:16:27 UTC from IEEE Xplore. Restrictions apply.