FIGURE 8. Testing involute gear design using the
3D printer and Solid Works.
1. Mechanical Components.
There are two routes to upgrading the legs.
I could attempt to beef up the legs in Solid Works
and continue to use the 3D printer plastic. The
other option is to change materials entirely, such
as aluminum. Considering the abuse the hexapod
undergoes, switching to aluminum seems like the
best option. The costs of getting CNC milled
aluminum with the current shape are certainly
much higher than printing plastic, but the reduction
of downtime at demos should more than make up
for the higher price tag. Some redesign will be in order
because a 3D printer can easily create any shape, but a
CNC mill takes knowledge of the machine’s capabilities
which must be accommodated for when designing.
2. Head Mechanism.
Ideally, a new head mechanism would be created
with metal parts. Considering that metal parts have much
less give compared to plastic, the parts cannot be worn
into each other as well as before. In order to solve the
problem, a properly designed gear must be fabricated. In
other words, replace the triangles with proper pitch angle
and circular involute curves for both the spur and bevel
gears. Fortunately, there are many resources for designing
proper, standard gears.
processing power for a
single servo motor. This results
in a communication bus increase from 1 Mbps to 3 Mbps.
It is also apparently possible to install a Linux kernel on
the Cortex M3.
Robotis recently announced and currently sells a new
motor: the MX- 28. There are some tremendous
advantages at only $20 more than the RX- 28.
Mechanically, the motor is similar to an RX- 28, but the
onboard controller is drastically different.
a. Magnetic Angle Sensor: The potentiometer has
been replaced with a magnet and a magnetic encoder:
the AS5045. Not only does this increase the resolution
from 1024 to 4096 and increase the effective servo
operation over the whole 360 degree range, but it is also
completely contact-less. This means incredible durability,
and it will solve all shaking issues that some of the RX- 28
motors were exhibiting.
b. PID Controller: The control system has been
upgraded to a full PID controller, following a very
commonly known control system. Being that all three
gains are adjustable, the motor can truly be properly
tuned for any system that can be modeled as a mass
spring damper. Also, the gains can be set to zero, so not
only is it a PID controller, but it could also be a P or PD
controller, or something non-standard like PI, D, or ID.
(I can’t wait to start tuning these motors!)
c. Microcontroller: Switching from the Atmega8, the
MX- 28 now features an ARM Cortex M3. That is a lot of
There will be new features implemented in the
hexapod program, including a better interface. Video feed
will be streamed wirelessly to a server so that users can
see how the face tracking is performing. The additional
features will, however, take more computation, so
considering that performance is also to be increased, a
newer computer will be in order.
Rather than use the z530 Intel Atom processor, I will
be implementing an E6XX series Atom processor. This
newer processor has onboard video encoding and
decoding, so vision processing and streaming should
operate very smoothly. Intel has developed a development
board named devboard which will mate nicely with a
fairly low profile Q7 platform, available from i Wave.
Some parts such as new motors are starting to arrive,
and I am busy cranking away at redesigning the next
generation of hexapod. You can see the shipment of
motors I just received in Figures 6 and 7. Figure 8 shows
properly designed spur and bevel gears using the involute
method. I will have an in-depth review of the MX- 28
motors, fabricate aluminum parts, describe the design of
involute gears, and recode and port my system to a
newer, faster platform in upcoming installments. SV
SERVO 12.2011 61