The story so far: The MowBot is still in an early stage of development. I have tracked down a highly sophisicated, inter-communicating, network-based programming environment called TCA . The primary reason I chose this environment is that it allows the user a great deal of flexibility in developing behavior-based control architectures. TCA developed at CMU's Learning Robot Lab. A shining example of the possibilities of TCA is Xavier, a sophisicated robot developed to navigate around an ever changing office environment to deliver messages, snap pictures, or to say something witty to the office's human occupants. Take a look for your self!

The robot will contain an on-board pentium-based PC running some varient of Linux. The ROACH project yielded a great deal of information about proximity sensors, and odometric correction that will help a great deal with the MowBot. The added processing power of the pentium allows for the addition of a vision-based navigation system employing the extremely inexpensive connectix quickcam, thanks to the folks on the Linux-Quickcam project this is now possible.

There are only two other autonamous lawn maintenance systems that I am aware of. One is a rather expensive, and not all that inteligent robot mower by the name of Weed Eater Hey, Epcot's got a bunch! You can pick one up at MondoTronics for about $2000 (if you have that kind of money laying around why not fund a far more interesting inteligent robotic system project (hint, hint ;-) The other system closely matches the base design of the MowBot, in more ways than one. This system is part of an ongoing systems engineering project at GMU. Unfortunately their page has little more than pictures of the robot available.

This project was intended as an extension to SNAP, to further the automation of my humble domicile in downtown San Jose, California. These ideas were simply another manifestation of my unwillingness to deal with the laborous duities of yard maintenance. (Not to mention an excellent opportunity to develop my robotic hardware and software development skills.)

These are the individual TCA modules that will compose the MowBot's control program.
 

P2P -Stereo vision program from the Stanford Vision Laboratory. This program uses two PGM images to generate disparity maps and estimated depth discontinuities. Note: the Makefile that comes in this package does not work as-is under Linux.