Results

Team name:	Cornflake
Number of gates:	11
Time:	1:16.97
Rank:	1st place

Technical Information

Chassis:	Commercial RC toy car (Audi TT)
Controller:	Parallax Javelin Java Stamp
Actuation:	Single motor driving all 4 wheels
	Standard digital servo for steering
	Standard hobby ESC motor driver
Vision:	CMUcam2 in color tracking mode communicating via serial interface
Odometry:	Single US Digital encoder attached to main shaft via a belt
	Quadrature signals converted to up/down ticks
	8-bit counter receives ticks and outputs to controller
	Theoretical precision: 0.1" linearly (1-D, assuming line followed exactly)
Speed control:	Simple (but effective) proportional velocity control

Pictures

• Top view: 1
• Guts: 1, 2, 3
• On course: 1, 2, 3, 4, 5, 6
• Decision point spy photos: 1, 2, 3
• Origin of the name: 1 (at one point we need to raise the camera and these worked quite well)

Movies
All the movies taken this year were embarrassing debugging videos. See last year's mobot for something similar. Note the lack of intelligent speed control and decision point logic.

Good Points

• Odometry (when line followed closely) very accurate
• Speed controller very effective
• CMUcam2's features made overall design very simple

Less Than Good Points

• Javelin is a toy: maximum serial port speed of 19.2Kbps, slow effective clock frequency
• 20fps update rate made steering wobbly and not robust
• CMUcam2 limits design options (no whole-frame vision algorithms possible)
• Method for decision point segment still relied heavily on noisy cmucam data

Future Plans

• More advanced vision system (perhaps a more standard camera and full-frame analysis)
• 2-D dead reckoning? (the line is only a suggestion...)
• Faster processor (to handle all of the above)
• Re-add gate sensor (useful ground truth data)

Other Information
We received a $1000 SURG grant for this project, for which Dr. Garth Zeglin kindly offered to serve as advisor. A complete report and budget sheet in fulfillment of the requirements of the grant will be submitted separately at the end of this semester. A significant portion of this SURG project is the design of a custom microcontroller board, which didn't get finished in time for the race. Schematics and a board layout will be presented at Meeting of the Minds along with the rest of the mobot report.

Links

• Official mobot site
• TCPulse's coverage of mobot race (a few good pictures one of me looking very disgusted)
• Jack's Mobot 2004 site (good overview of the tracking algorithm)
• Jack's (incomplete) Mobot 2005 site

Last modified: 4/19/2005