Approach and Overall DesignThe RoboMagellan competitions are
useful for remaining focused on a project that
spans years. The platform is really intended
for robotics R&D and as such is designed for
fairly rapid modification. Images of the
machine will show it in different
configurations, most noticable being the positioning
of the sonar transducers. The budget for this project is quite
modest. There are no laser scanners or industrial
grade IMU devices to be found on-board. Intrepid's chassis is that of a Traxxas E-Maxx 4WD R/C vehicle. The motor controller was
replaced with twin Parallax HB-25 motor drivers
feeding the stock 550 'Titan' motors. This
combination has worked quite well. Intrepid
doesn't currently move very fast, but this has
nothing to do with the drive system and everything
to do with safety. As the obstacle avoidance
techniques mature the speed will also
increase. Intrepid can move fast enough that you
would need to run flat out to keep up! There are several microprocessors onboard,
including a BDMICRO MAVRIC-IIB Atmel ATMega128
controller, used to manage the sensors, compass,
IMU and drive system. A Pentium-M mini-ITX board running MATLAB is the mission controller and image
processor. The main processor communicates to
the embedded controllers via serial link. Some
people have been puzzled by the choice of
using MATLAB. There are two main reasons for
this. 1) Image processing is a high priority
for this project. 2) This approach allows for
rapid prototyping of ideas and
algorithms. Imaging requires computational
power, there's no getting around it. This is
the reason for the mini-ITX which is certainly
a bit bulky for the E-Maxx, however having
lots of computational power has its
advantages. We are able to leverage the
scripting power (no re-compilation) as well as
the image processing library wihich is packed
with routines that don't have to be
reinvented. One may then concentrate on the
application at hand. This powerful combination
has only barely been tapped up to this
point. Current areas of research are mapping
and path planning. Software development
happens right on the robot. The IMU is a custom
gyro-accelerometer-magnetometer set based on an ATMega32
processor. Intrepid uses both infrared and sonar sensors for object
avoidance, and uses a camera to identify and navigate to cone targets
at way points. It also has a GPS receiver used to correct for odometry
drift. Imaging acquisition is accomplished using a standard USB
connected CCD webcam. Intrepid also has a WiFi data link that allows
for easy remote testing and data capture. | 
