This progress report details significant advancements in the design of a low-cost humanoid robotic testbed, focusing on high-fidelity CAD models of the robot's hips and legs, and updates on actuator selection. The team resolved mechanical power transmission issues by opting for belt and pulley systems and successfully created a functional MATLAB model for robot locomotion. Future work aims to finalize CAD models, optimize MATLAB simulations, and prepare for an upcoming oral presentation.

1. Design of a Low-Cost Humanoid Robotic Testbed for Open Source Educational
Development
Progress Report #2
AME 441aL
Tyler Presser, Nicholas Ilibasic, Elliott Hoppe, William Alpert
Group 13
USC Aerospace and Mechanical Engineering
October 9, 2020

2. Abstract
Since Progress Report #1, significant development has been achieved on RoBOT’s design. The
primary focus of the last three weeks has been to design high-fidelity CAD models of RoBOT’s
hips and legs as well as increase fidelity in the MATLAB and Simulink simulations. Using
knowledge gained through research and preliminary research in prior weeks, the team worked to
integrate different mechanical components such as linkages and joints to analyze design feasibility
in great detail. This report will briefly explain the specific details of the progress achieved in the
CAD models and simulation, as well as address updates to the timeline and plans for the coming
weeks.
High-Fidelity CAD Designs
At this point, the team has achieved a high-fidelity CAD design of RoBOT’s hips, femur, knee,
shin, ankle, and foot. The only part of the design that still needs to be finished in CAD is the central
housing that will sit between the two hip joints and store the electronics. This design depends on
the hips and legs, which drive the movement of RoBOT, so it was not the main priority for this
progress report. However, the central housing high-fidelity CAD model will be completed in the
next week. An isometric view of RoBOT’s hip and leg assembly is shown below in Figure 1.
Figure 1. Isometric view of fully assembled left leg. This angle depicts the inner portion of the
leg.
As seen in the image, motion of the legs is driven by two AK10-9 motors at the hip and one AK80-
9 motor at the knee joint. The rationale behind the selection of these two types of motors will be
provided in the next section titled “Selection of Actuators”. Both the femur and shin will be made
of carbon fiber tubing with a width of 24 mm. The joint connection pieces (shown in red) and the

3. feet (shown in yellow) will be 3D-printed with Carbon Fiber PLA to fit RoBOT’s design
requirements. A more detailed description of the mechanics of the hips and legs will be described
in further detail below.
One of the major problems that has been solved since the previous progress report was how
mechanical power will be driven from each actuator to achieve rotation at the hip, knee, and ankle
joints. The team was deciding between rigid linkages and belts, but ultimately decided to use belts
mounted on pulleys in both the femur and the shin.
In order to drive power at the ankle joint, a belt and pulley system will be used. This belt will run
parallel to the shin, but outside of the tubing as seen in Figure 2 below. Power will be driven in the
pulley below the knee from the AK80-9 motor, which is mounted to the circular red joint. This
belt connects to the other pulley located at the ankle joint, which will drive rotation in the foot.
The design of the knee joint was relatively simple because there was only one motor on the axis
of rotation.
Figure 2. Isometric view of RoBOT’s right knee, shin, ankle, and foot. Belts will run along the
pulleys attached to this assembly but are not included in this image.
The more challenging aspect of the leg design was the hip joint, which involves two AK10-9
motors mounted coaxially on opposite sides of the joint. As seen in a more detailed view in Figure
3, the motor located on the inside of the joint is mounted to the central housing and its output is
responsible for rotation of the femur linkage about the hip joint. The second motor is mounted

4. directly to the hip casing on the opposite side. However, its output drives the rotation of a pulley
inside the hip joint that moves a belt. This belt runs from the hip joint through the carbon fiber
tubing of the femur and is attached to another pulley at the knee. This belt drives rotation at the
knee joint, which can move the shin linkage.
Figure 3. Back view of RoBOT’s right hip joint. The hip casing has been shown with
transparency to provide clarity on how the system is connected. Belts will run along the pulleys
attached to this assembly but are not included in this image.
Since the team has decided to go with a design that uses belts to drive mechanical power, it is
important that there is a way for belt tension to be adjusted. To adjust tension in the belt that runs
from the knee to ankle, a third pulley is mounted to the knee joint near the other pulley. This pulley
can move linearly through a slotted box with a set screw and nut at the back and can be seen in
Figure 4 below.

5. Figure 4. Detailed view of RoBOT’s adjustable pulley near the right knee joint. The hip casing
has been shown with transparency to provide clarity on how the system is connected. Belts will
run along the pulleys attached to this assembly but are not included in this image.
For the belt that runs from the hip joint to the knee joint, a different tensioning system was
implemented that allows access to the belt inside the tubing. The white tab shown in Figure 5 can
slide out of the assembly so that a user can access the belt within the casing. The roller (shown as
a black cylinder) can be adjusted by a set screw and nut to increase tension. It does this by pushing
one side of the belt down, thus increasing the total length of the belt. In Figure 5, the path of the
belt is shown with yellow lines. It is important to note the entire belt will run on one side of the
roller, not around it.

6. Figure 5. Detailed view of RoBOT’s tensioning system below the right hip joint. The hip casing
has been shown with transparency to provide clarity on how the system is connected. Belts are
shown as yellow lines and are included for illustrative purposes.
The next steps to be completed in NX are the primary FEA analysis and motion simulations for
the robot’s legs.
Selection of Actuators
In order to determine which motors should be used in each joint, calculations based on MATLAB
simulations and the LIPM models mentioned in PR1 were performed. As a result, two different
actuators were selected to meet the needs of RoBOT’s design: AK10-9 and AK80-9. The AK10-
9 is the stronger of the two with a rated torque of 18 Nm and weight of 820 grams. It will be used
to rotate the femur linkage as well as the knee. Since the ankle torque requirements are much less,
power will be supplied by the AK80-9 motor, which has a rated torque of 9 Nm and weight of 485
grams. It is also cheaper at a cost of $580 instead of $699 for the AK10-9. Both of these motors
make use of a planetary gearbox in which the input shaft and output shaft are aligned, allowing for
a high torque density.
The math used to calculate the actuator choices can also be validated by comparing RoBOT’s
actuator choice to existing robots. One such robot is the Archie Robot from the Vienna Institute of
mechanics and Mechatronics [3].

7. With an approximate weight of about 3.5kg, using ZMP simulations for stability and the LIPM
modelw for motion they found that their maximum required torque in the knee joint for Archie is
12Nm.
Figure 6. Arhcie bipedal robot from Vienna.
MATLAB/Simulink Simulation
At the time of writing the team as generated a functional MATLAB model of the RoBOT walking,
including a custom generated trajectory designed around the dimensions of the robot. This is of
course only a first step towards a final representation of the locomotion of RoBOT, however it is
the most significant threshold to cross in the process of simulation since the future of this
simulation will consist primarily of optimization. A time series of RoBOT walking in this
simulation is shown below.
Figure 7: RoBOT’s uneasy first steps in the MATLAB simulation display a significant step but
also makes clear a number of parameters that will require further optimization.

8. As discussed in PR1, the basic simulation model that is being used to develop RoBOT is based off
of an open source repository designed by Sebastian Castro and released by MathWorks to aid in
the development of bipedal robots [2]. The current simulation runs a model of RoBOT directly in
the open-source framework while a more detailed, independent framework is created specifically
to run simulations for RoBOT. The key reason a new framework must be created is that RoBOT
is designed around purely dynamic walking and as such is not fitted with an ankle-roll actuator.
This is done to reduce leg inertia and cost and should result in a cheaper and more agile and
responsive robot. Since the presence of an ankle roll joint is integral to the functioning of the
MathWorks framework, a specially designed simulation framework must be built from scratch.
It should be noted that the project began with an intention to create the higher-fidelity simulation
in a co-simulation environment with Siemens NX, however further investigation into the feasibility
of this plan has revealed that this difficult if not impossible. As a substitute, a similarly high-
fidelity model will be developed in Simulink and MATLAB. An eventual goal for the simulation
is to include the CAD of the robot to most accurately represent the inertial properties of the links,
however this more detailed simulation is constrained by available RAM and may not be possible
until local access to USC computing resources is available again.
Timeline Update and Project Setbacks
Initial CAD designs with increased fidelity have been created for the legs and hips of RoBOT. The
actuator selection based on more detailed hand calculation of the torques required to raise and
lower the robot under the power of a single leg. Up until this point, a greater focus has been placed
on the leg/hip design and simulation over the central housing design, which is still in progress. At
the time of the Progress Report #1, the team expected to have the central housing design complete
by this progress report. However, an extra week has been allotted for this task since the team has
realized that its design is driven by the legs and hips and only serves to house the electronics of
RoBOT. This should be a fairly simple task that can be completed by next week. The simulation
section of the timeline has been updated to reflect the setback of the incompatibility of Siemens
NX and Simulink for co-simulation. The timeline now reflects the pivot to more detailed
simulation all within Simulink and the creation of a modified framework to support that simulation.
Another update to the timeline is a section to plan for the Oral Presentation, including a rehearsal
with Dr. Staelens on October 19th
and a final delivery on October 26th
.
Future Work
The major focus over the next three weeks will be to finalize the CAD models of the central
housing, hips, and legs. A complete CAD model should be ready to go for the last progress report.
The MATLAB simulation shown above opens a series of doors for optimization of controls,
electromechanical systems, and even geometry. Future work for the MATLAB simulation also
will center around finding our power and torque margins based on our chosen actuators, and set
minimum torque values determined earlier in this report.
The CAD for the two-leg links has been carefully parametrized in order to allow for this
functionality. The first result from the simulation is that the tendency of a high inertia leg to cause
the entire robot to yaw is extremely detrimental to performance, and that every effort must be made
to reduce the mass of components placed low on the leg. Outside of the Simulink model, further

9. mechanical analysis will also be done to verify that all loads can be supported by the linkages and
pins in RoBOT. Finite Element Analysis will be conducted on linkages and shear calculations will
be performed on pins. This analysis will drive some of the final decisions such as material choice.
As these goals are worked towards, the team will also begin to prepare for the oral presentation,
which is on October 26th
. The team will start to create slides next week to have ready for a rehearsal
with Dr. Staelens on October 19th
.
References
[1] S. Kajita, F. Kanehiro, K. Kaneko, K. Yokoi and H. Hirukawa, "The 3D linear inverted
pendulum mode: a simple modeling for a biped walking pattern generation," Proceedings 2001
IEEE/RSJ International Conference on Intelligent Robots and Systems. Expanding the Societal
Role of Robotics in the Next Millennium (Cat. No.01CH37180), Maui, HI, USA, 2001, pp. 239-
246 vol.1.
[2] MathWorks Student Competitions Team (2020). MATLAB and Simulink Robotics Arena:
Walking Robot (https://github.com/mathworks-robotics/msra-walking-robot), GitHub.
Retrieved September 17, 2020.
[3]BajramiXh.,etal.”Kinematics and dynamics modelling of the biped robot.” IFAC Proceedings
Volumes 46.8,69-73,2013

10. Appendix A: Updated Cost and Weight Estimate

11. Appendix B: Updated Timetable