2009 Electronics, Robotics and Automotive Mechanics Conference

AxeBot Robot: The Mechanical Design for an Autonomous Omnidirectional

Mobile Robot

Tiago P. do Nascimento, Augusto Loureiro da Costa, Cristiane Correa Paim

Post-graduation Program in Electrical Engineering
Universidade Federal da Bahia
Salvador, Bahia, Brasil
tiagopn@ieee.org, augusto.loureiro@ufba.br, cpaim@ufba.br

Abstract algorithms, the goal of RoboCup is to build a team of

humanoid robots that can beat the current World Cup
The AxeBot robot’s mechanical design, a fully champions by the year 2050.
autonomous mobile robot, for the RoboCup Small Size The AxeBot uses three omnidirectional wheels,
League, is presented in this paper. The AxeBot robot positioned on a circle with an angle of 120o among
uses three omnidirectional wheels for movement and is each wheel, to move in different directions. Three
equipped by a shooting device for shooting the ball in Maxxon A-22 motors are used to drive the
different directions. Once the AxeBot robot is a fully omnidirectional wheels, one motor per wheel. These
autonomous mobile robot all the sensors, engines, motors are controlled by two Brainstem Moto 1.0 and a
servos, batteries, and the computer system, must be cascade controller made to control the robot trajectory
embedded on. The project can be separated in four [2] [3]. The AxeBot also holds a shooting device to
different parts: the chassis design, the wheel design, kick the ball in different directions, a Vision System
the shooting device design and the overall assembly with a CMUCam Plus and GP202 Infra-red sensor [4],
which makes a shell design possible to cover the whole a embedded Computer System based on StrongArm,
robot. The AxeBot mechanical design brings up a new called StarGate Kit and a IEEE 802.11 wireless
chassis concept for three wheels omnidirectional network card. This work presents the mechanical
robot, also present a new shooting device, and finally project to enclose these equipments into an fully
present AxeBots prototype assembly. autonomous omnidirectional robot called AxeBot. The
complete AxeBot dynamics and kinematics model can
1. Introduction be found in [5], this model was used to specify some
mechanical parameter, like the wheel diameter.
The RoboCup Initiative is an international research
group whose aims are to promote the fields of Robotics 2. The Chassis
and Artificial Intelligence. A standard challenge, a
soccer match performed by autonomous robot teams, The chassis of the robot is the frame to which all
was proposed in 1996 [1]. Initially with three different other components can be attached, directly or
leagues 2D: Robot Soccer Simulation league, Small indirectly. Therefore the chassis must be strong enough
Size Robot league, and Middle Size Robot league. to carry the weight of all parts when the robot is in rest
Nowadays these leagues have been increased up to: o in movement. The chassis has to withstand the forces
Four-Legged League, Humanoid League, Middle Size on it, caused by the acceleration of the robot as well.
League, RoboCup Junior Soccer, Small Size League, Another important requirement of the chassis is that it
Soccer Simulation, Standard Robot League. Also, fixes all components in a stiff way, so that there will be
another challenge, the RoboCup Rescue was proposed small relative displacements of the components within
in 1999 to show that the result from the robot soccer the robot, during acceleration and deceleration. This is
research could be directly applied on a real world particular important for the three driving motors, which
problem like a disaster rescue made by robots. Through are positioned on the ground plane with an angle of
the integration of technology and advanced computer 120o between each motor. The performance of the

978-0-7695-3799-3/09 $26.00 © 2009 IEEE 159

187
DOI 10.1109/CERMA.2009.77
control of the robot is dependent on a precise and stiff This sandwich construction gives the whole chassis
placement of the motors [6]. The chassis has to be more stiffness, and so the total thickness of both the
strong enough also to withstand a collision of the robot chassis can probably be lower than using one chassis
against the wall or against another robot, with the part.
highest possible impact velocity that can occur. Finally
the chassis has to be built with the smallest amount of
material. At first to reduce the costs, and to minimize
the total weight of the robot. Less weight requires less
power to accelerate. So with the same motors, less
weight gives you more acceleration. This is of course
only true, when all the power generated by the motors
can be transferred, via the wheels, to the ground. In
other words, the wheels must have enough traction that Figure 1: Lower chassis
there will be no slip between the wheels and the ground
[7].

2.1. Material

Fiberglass was used to build the chassis. This choice

is purely financial, because the material is not
expensive (although it is strong) and there is no need to
hire a professional constructor. The building of all the
chassis (six in total) can be done by the team members Figure 2: Upper and lower chassis attached to each
themselves. Only the moulds have to be built by a other
professional. The upper and lower chassis can be made
using one mould that can be adjusted to produce the 2.3. Chassis mould
different chassis.
To build these parts, a mould was made. This is just
2.2. Design a negative of the actual robot parts. In figure 3 the
mould of the upper chassis. To change this mould in
The primary goal of the design is the fixation of the the mould for the lower chassis, where the ground plate
motors in the desired positions. Therefore a ground does not have holes, the indicated pieces (with white
plate with 3 slots for the motors is modeled. At the stripes) and the not indicated left piece (symmetric to
front side each motor can be attached to the chassis. At the most right part) should be lowered 4 mm. For the
the rim of the ground plate an edge is attached to give upper and lower chassis, the basis mould is exactly the
the chassis more torsion stiffness. This edge can also be same. Only piece one and two are different for the two
used for attaching other components of the robot, like chassis, the motor piece and the shooting system piece
the covering shell. Also there is a cutout to create space are the same.
for the shooting device of the robot. In section 5 the
design of this device will be discussed. However no
final design will be presented and therefore we stick
with this assumption that the shooting device needs
these cutouts. All edges are rounded, because this will
make the construction of the easier part. The final part,
the lower chassis, is shown in the figure below. This
part is modeled in Solid Edge.
To get a stiffer and stronger chassis, a second
chassis part, the upper chassis, is modeled. This is Figure 3: Chassis mould
almost an exact copy of the first part, only now there
are 3 cutouts that provide more space for placing the 3. Wheels
components of the robot. These cutouts also save some
material and therefore weight. The both parts are The AxeBot robot is equipped with three wheels
attached to each other as shown in the picture below. positioned on a circle with an angle of 120o among

188
160
each wheel. These wheels have to enable the for maximum grip. This however ruined the carpet and
omnidirectionality of the AxeBot robot. This means rubber rings were added in the design to obtain
that the wheels have to be able to let the robot make maximum grip without ruining the carpet. The rings are
two translational movements (in x and y-direction, see circular with a circular profile.
figure 4) without rotating the robot around its z-axis
(the axis perpendicular to the y and x-axis, that is
rotation in figure 4). The wheels have also to enable a
rotation of the total robot around the z-axis.

Figure 5: Exploded view of the wheel

Therefore the rubber rings are also used for the

AxeBot 2006. Rubber rings can be bought in several
sizes and since they are highly elastic it wasn’t difficult
to find a ring of a right size. Since there are more than
Figure 4: AxeBot wheels positions one ‘right sizes’ and the geometry of the ring is that
simple, no technical drawing of the ring was made.
Nevertheless, the wheels have to be as small and
light as possible to minimize weight and moment of 3.2. Roller
inertia but still remain usable and manageable. The
wheels are based on an existing design of an The geometry of the rollers may not restrict the
omnidirectional wheel from the Cornell Robot 2003 rotation of the roller and should enable the placement
[8]. Figure 5 shows an exploded view the final version of the rubber ring, without the ring falling off. This can
of a wheel. be easily obtained (see drawings). The only problem is
The two shells are connected to each other by friction with their axles and with the shells.
screws and hold every part on the right place. The hub The geometry of the rollers can influence the
is also attached to the shells by screws. The hub is friction with the shells and the friction with its roller
mounted on the output axle of a motor by a screw to axle. To minimize the possibility of wear on the contact
transfer the rotational output of a motor to the wheel. area between the roller and the roller axle, this contact
The rings of the rollers are in contact with the floor. A area should be as big as possible.
roller can rotate around its roller axle. When the torque of the motor is transferred through
As mentioned above, the wheel has to enable two a roller to the ground (driving the robot), the roller that
translations (x and y, see figure 4) without rotating is on the ground is pressed against the shells. It will
around its z-axis. The whole wheel ensures one occur often that, in this situation, the desired driving
translation by rotating around the output shaft of the direction also requires a rotation of the roller (then the
motors while the rollers ensure the other translation. summation of the two translational movements will
Combining these translations on a proper way a robot results in the desired driving direction). The rotating
can move anywhere in a plane or make a rotation. roller is, in this situation, pressed against the shells
which, in some cases, can result in wear of the roller
3.1. Ring and or the shells. This depends on the material of the
roller, the material of the shell, the magnitude of the
The ring is the only part of the wheel that is in force which presses the surfaces on each other (in this
contact with the floor. Note that a wheel can also be in case it is the torque of the motor) and the geometry of
contact with the floor by two rings. To obtain both contact surfaces. Only the materials and geometry
maximum grip (no slip) the Cornell Robot 2003 team can be chosen in the design process of the wheels.
first developed rollers without rings. The rollers had A small contact area results in low friction but
sharp edges to cut into the carpet of the football field possible wear of one of the two surfaces and though

189
161
materials have to be used to avoid wear. A large steel axles since the friction-coefficient between steel
contact area will not cause wear but will result in a and steel is lower than that between steel and
large friction force. An optimum for the contact area aluminum. A lubricant can also be used to even more
and the materials has to be found. The geometry has to reduce friction.
be machinable also.
Problems that are mentioned above did not occur 3.3. Roller axle
with the design of the rollers of the Cornell Robot
2003. Therefore the same geometry for the rollers is As mentioned above, the use of Delrin for the roller
used for the AxeBot. axles was investigated since it would reduce the wear
The Cornell Robot 2003 team designed a wheel with of the aluminum rollers. In a static situation was
15 rollers that worked very well. Therefore also 15 calculated whether Delrin axles of 2.4 mm diameter
rollers are used in each wheel of the AxeBot 2006. would be strong enough. This was also done in a
The prototypes of the wheels of the Cornell Robot dynamical situation (dropping the robot on the floor
2003 first had Delrin rollers to minimize friction and and landing on one roller), but without using a Finite
weight. Delrin is a kind of plastic which is used with Element Method this did not result in realistic results.
moving contact surfaces because of its low friction When the total weight of 3.5 kg of one AxeBot 2006
coefficients with other materials. After a few test with would completely be on one roller axle this would
the prototype robot it became clear that the Delrin- result in a shear stress in the axle.
rollers easily broke with a collision. Therefore the In this situation the shear stress can be calculated by
Cornell Robot 2003 was equipped with aluminum dividing the force on the axle (due to the weight of the
rollers. These were strong enough to withstand AxeBot) by the area of the shear plane. The area of the
collisions and aluminum has a low density (compared shear plane is of course the area of a circle with a
to other metals). radius similar to the radius of the axles. Note that the
However, during other prototype tests of the Cornell weight of the AxeBot has to be divided by two since
Robot 2003 team some aluminum residue built up on there are two shear planes in one axle.
the steel rollers axles. The Cornell Robot 2003 did not The magnitude of the shear stress would be 3.8
encounter problems due to the wearing of the MPa. Delrin starts to plastically deform in due to shear
aluminum rollers, but to optimize the design of the at around 44 MPa. Statically, Delrin axles would be
AxeBot wheels this problem was solved. strong enough.
To avoid wearing of the aluminum rollers other However, this calculation was not necessary it
material for the axles can be used or the rollers can be became clear that it is not possible to produce Delrin
made of a tougher material. bars of 2.5 mm (diameter). Therefore steel axles will be
After a few calculations it became clear that roller used (aluminum axles would result in more friction).
axles of Delrin (to reduce the friction) are strong To reduce friction, the axles were coated with a
enough (see the section about the axles), but it is not lubricant like carbon. Lubricants like carbon are easily
possible to produce thin bars of Delrin. Therefore steel available.
axles are used, the same material as the roller axles of The Cornell robot team 2003 documentation does
the Cornell Robot 2003. not mention problems of wear of their polycarbonate
So to avoid wear of the rollers a more though shells due to their steel roller axles. As one can be read
material than aluminum has to be used for the rollers. further down this section, the shells of the AxeBot
Steel is more though and an easily obtainable and wheels will be made of aluminum or polycarbonate.
cheap material. A disadvantage of steel compared to The coefficient of friction between aluminum (shells)
aluminum is its higher density. This will increase the and steel (axles) and between polycarbonate (shells)
moment of the inertia which ‘costs’ more torque of the and steel (axles) are of equal sizes (about 0.45 [-]).
motor. The total moment of inertia of a wheel with Also, the geometry of the wheels of the Cornell robot
steel rollers is 1.39×10−4kgm2 and the moment of and the AxeBot are almost the same. Therefore it is
inertia of a wheel with aluminum rollers is most likely that wearing due to friction will not be a
1.30×10−4kgm2. Using steel rollers instead of problem with steel roller axles and polycarbonate or
aluminum rollers would increase the moment of inertia aluminum shells.
by 7 this increase is neglectable small and steel rollers Production the axles can easily be made out of a
can be used. steel bar. Depending on the available diameters of steel
Concerning friction, using steel rollers and steel bars, the diameter of the axles could be adjusted. The
axles is also better than using aluminum rollers and edges of the axles are rounded to avoid sharp corners.

190
162
3.4. Hub then with one connection point and smaller reaction
forces will act on the connection points.
The hub connects the wheel to the axle of the motor.
A M2-bolt can be used for this purpose, the hole in the
hub where this bolt will be placed is dimensioned 1.5
mm in diameter so screw thread can be made to fit in
the M2-bolt. The hub is connected to the shells by three
M3-screws. Corresponding holes of 2.5 mm will be
made in the hub and the shells. The geometry of the
hub can be changed to facilitate the production process.
The hub can be made out of aluminum or Figure 6: The AxeBot shooting device
polycarbonate, both light materials. To investigate the
option (and price) of injection moulding the hub out of 5. Shell and AxeBot Assembly
polycarbonate a sketch of a design for a mould has
been made. The shell is the cover of the total robot. It will be
made from fiberglass as well, for the same reasons as
3.5. Shells stated in the section about the chassis. In the figure
below the design of the shell is shown. There are
Almost the all geometry of the shells of the Cornell cutouts to make room for the wheels as well as for the
Robot 2003 is used. Only little changes in the slots for shooting device. The diameter of the shell is 178 mm.
the rollers of the axles have been made to facilitate the Because the maximum allowable diameter is 180 mm, a
production process of the shells. In the section about margin of 2 mm is created.
the production of the shells a different design for the On top of the shell a vision system will be mounted.
slots of the axles is presented to make the production of For the sake of compactness of the robot the height
the shells easily. should be as small as possible, with all the parts fitting
The shells can be made out of aluminum or a in. This mould can be built with the simple milling
though, light plastic like polycarbonate. The Cornell machine that is available.
Robot had shells of polycarbonate. The best To assemble the robot the most important demand is
machinable material is preferred, and the one used on to obtain a total centre of gravity that has the lowest
the AxeBot’s shell. When using a plastic that can be possible position in the robot. This will give the robot
injection molded, molds have been designed to check positive driving abilities. Another demand is of course
the prices of injection moulding. that all the parts fit in the maximum height of 150 mm.
To check this, all the parts of the robot have been
4. Shooting Device modeled in Solid Edge. In figure 7 an exploded view of
the total robot assembly is shown.
This design consists of a vertical arm (the kick arm), The robot consists of an upper and lower chassis.
that can swing around an axle which is fixed to the Three motors with three omnidirectional wheels are
robot. This movement will be actuated by one of the attached to it. On the bottom of the robot, the two
two servos, the kicking servo, that is also fixed to the battery packs are placed, because these parts have the
robot. The other servo, the directional servo, is largest mass. The three motor processors are attached
attached to the bottom of the kick arm (the to two general processors. Also the overall processor
servosocket). A pendulum-like system is formed, so a with the data-transfer-unit is assembled. The latter parts
large mass is concentrated in the lower part of a are assembled in a way that is most compact. With this
rotating arm. The kicking servo has a kicking plate assembly it is possible for all parts to fit in a shell with
attached to it which can rotate and thereby makes it height of 100 mm.
possible to shoot in different directions. The kicking A model for the shell of the robot is designed. Also
plate can be positioned very accurately since servos are a molt to produce these parts is modeled. It is possible
designed for these kinds of tasks. The kicking plate is to produce this model by using the milling machine that
also connected to the servosocket. Therefore the is available at the university. The assembly of all the
collision force between the ball and the kicking plate parts, except the shooting device, shows that a shell
will be guided to and divided by these two connection height of 100 mm is possible.
points. There will be less bending in the kicking plate

191
163
 will support our research in to multi-robot systems, as
it can be seen in figure 9 below.

Figure 9: AxeBot photo

Figure 7: Exploded view of the total robot assembly 7. References

[1] Kitano, H. “Robocup: The robot world cup initiative”. In:
Proc. of The First International Conference on Autonomous
Agent (Agents-97)). Marina del Ray, The ACM Press. 1997.

[2] Franco, A. C. “Geração e controle de trajetória de robôs

móveis omni-direcionais”, Master’s thesis, Programa de Pós-
Graduação Mecatrônica, UFBA. 2007.

[3] PIRES, E. J. S.; MACHADO, J. A. T.; OLIVEIRA, P. B.

de M. “Robot trajectory planning using multi-objective
genetic algorithm optimization”. In: DEB, K. et al. (Ed.).
GECCO (1). [S.l.]: Springer, 2004. (Lecture Notes in
Computer Science, v. 3102), p.615-626. ISBN 3-540-22344-
Figure 8: Total robot assembly 4.

6. Conclusion [4] Oliveira, L. R., Costa, A. L., Schnitman, L. and Souza, J.

“An architecture of sensor fusion for spatial location of
The AxeBot mechanical design, a fully autonomous objects in mobile robotics”, in B. H. Spring-Verlag (ed.),
mobile robot, for the RoboCup Small Size League, was Encontro Português de Inteligência Artificial, EPIA’2005,
Covilhã, 2005. pp. 462–473.
presented in this paper. This mechanical design brings
up a new concept of chassis for three wheels [5] Nascimento, T. P., Costa, A. L., Paim, C. C. “Uma
omnidirectional mobile for RoboCup F-180 league, abordagem multivariável para modelagem de robôs móveis
that can be built easily and cheap. Also a new omnidirecionais”, XVI CBA - Congresso Brasileiro de
effectuator mobile robot design for RoboCup F-180 Automação. 2008.
league is presented here. This new effectuator allows
the mobile robot to shoot the ball in different [6] ANGELES, J. “Fundamentals of Robotic Mechanical
directions, instead of just shoot a head like the other Systems: Theory, Methods, and Algorithms”. 2. ed. New
shooting devices. Finally the mechanical project York: Springer-Verlag New York, Inc., 2003.
presented here encloses all part, sensor, actuators, [7] B. Carter, M. Good, M. Dorohoff, J. Lew, R. L. W. II,
effectuator, computer systems, wheels, chassis and and P. Gallina, “Mechanical design and modeling of an
cover shell into the AxeBot prototype. The AxeBot omni-directional robocup player,” in Proceedings RoboCup
robot was concept for academical proposes, using the 2001 International Symposium, (Athens, Ohio), pp. 1–10,
robot soccer as a laboratory to research in Autonomous 2001.
Mobile Robots, Artificial Intelligence and related
areas. Looking forward, in a few months, four more [8] Anderson, G., Chang, C., Chung, D., Evansic, L., Law,
AxeBots are expected to be built like the two in figure H., Richardson, S., Robers, J., Sterk, K. and Yim, J. 2003
cornell robocup, mechanical group final documentation,
9. These robots will form the MecaTeam F-180, and
Technical report, http://robocup.mae.cornell.edu/. 2003.

192
164