Introduction
Tarzan is an outdoor robot. His habitat ranges from cityscape to
semi-improved land. His purpose is simply to provide a learning
experience for realistic field robotics. In the Spring & Summer of
2008, he will "compete" in several RoboMagellenish events with very
simple reactive behaviors and minimal sensors. Later, I'll work on
improvements to sonar ranging, 3-D vision and a bit more intelligence,
as progress continues in discovering the art of creating an autonomous
creature.
Numerous changes have occurred along the way, but the overall goal of
the project remains the same, which is to auto-navigate over varied
terrain in places people would be comfortable at about 10 miles per
hour for maybe an hour or two.
5/14/08 DPRG Outdoor Challenge
Tarzan and I traveled this past weekend to Dallas to enter the
internationally famous DPRG Outdoor Challenge. The Texas hospitality
put the shy jungle youth at ease and he performed well, better than
expected.
Since last update, the PC was replaced by a 5V, 5W fitPC, again
lowering the power consumption of this boat anchor processor. As you
can see in the picture, it's all fitting better on the deck and I'm
happier with the package now. I rededicated one of the 4 Lipo packs to
the fitPC, so there are 3 now for the motor.
Also since last update, power distribution was completely redone with
separate boxes for the electronics and motor. It is now easy to
disconnect the batteries. Finally, an auxiliary DC source can power the
electronics while the batteries are kept fresh standing by or even
charged in parallel, so it's no longer necessary to power down the PC
while charging. Yay!
Notice the sonar board with a single Maxbotix unit. This was what won
event #3 in Dallas with absolutely no ground testing. I'll have much
more to say about sonar soon.
4/10/08 All ready for first ground-testing
The behavioral program & speed control now runs on a VIA Artigo
pico-ITX 1GHz x86, which is the black box on the rear deck above, while
the sensor/actuator interface is a single Atmega128 (the green board on
the fore deck above) connected to the PC by RS232/USB. The role of the
PC on my robots is in doubt and will likely disappear for the next
version, to be replaced by a single board containing everything needed
for both hardware interfacing and behavior. I may shift to this single
board architecture soon after the upcoming PARTS Outdoor Challenge May
25. We'll see a bit more about how best to do this at the presentation
by Pat Nystrom at the May PARTS Mtg, where he'll discuss mating a cheap
32b processor such as Cortex ARMx with a several thousand gate FPGA.
Testing of Tarzan on grass has gone well. He can definitely move
effectively between waypoints. Fine turning is ok but jack-turning is
still an ordeal to watch due to the poor turning radius. However, in
reality, it remains to be seen how important the need for tight radius
turns will be.
The Artigo is powered by 12VDC, which currently comes from either a
wall switcher or a DC SPS powered by an 8-cell Lipo pack. At this point
I still haven't built a circuit to allow simultaneous powering from
wall and battery. Consequently, by leaving the battery pack connected
to the unused DC switcher for several days during development on the
bench, I've run the battery pack dead several times now. What a pain.
This highlights another issue in constructing robots, ie the
awkwardness of power supplies and the time needed to create a
multi-source supply and isolation between electronics and motor
functions. Some refinement on how to do this efficiently is in order
and it's important to consider incorporating the supply hub and
terminals onto the next controller board for a quicker setup. This
would allow plugging in of each battery to the board and each remotely
mounted power switch.
About the author...
I'm a retired EE/exec from the IC industry (primarily Maxim Integrated Products, Inc.), currently the president of