2005 RoboMagellan for RoboGames

Team Members: Kallahar, Kelvin, Echo, Tacitus, Lorin

About three weeks before the 2005 RoboGames we decided to enter a bot in the RoboMagellan contest. The goal of the event is to have your autonomous vehicle traverse an outdoor course without any human control. This event is far more interesting to us than, say, battlebots because they're really just R/C cars.

Our base platform is a 4wd Radio Shack something or other we bought from a friend for $20. We modded it by removing the electronics, replacing their steering servo with our own, and trippling the battery capacity. For the last week before the event, we were up till 3am every day (and to our day jobs by 8am the next day) trying to get everything done and working. Anyone who has worked on a crunch project knows that mistakes are made, and minor problems turn into big problems when you're tired. More than once it took an hour to figure out something that should have been figured out in minutes.

The original design was a Via Epia800 mini-itx in a custom lexan case. Unfortunately, the power needed to run it was more than our 4Ah sealed lead acid batteries could reliably supply. The power draw was causing the batteries to drop below the 10.5V that the mini-itx needs to run, causing the computer to reboot. Twelve hours before the race, we made the decision to switch to my laptop, which meant fabricating a new mounting plate for all the electronics. Fortunately, Lorin had an aluminum tray sitting around which fit perfectly!

We got about 5 hours of sleep that night, then headed for SFSU for the 9:30am start in the quad. We were the first ones there so we set up to do some testing and final tweaking. Unfortunately, things started to go wrong there. For the future: key i2c connectors, connecting them backwards gives random bad data! Fortunately the people running the event were gracious and gave everyone time to do some final tweaking. We made all three of our runs by 1pm, but were unfortunately unable to do well on the course. On the plus side, no one was able to complete the entire course, and I did an interview for TV :)

For the next event:

• Low power computer, such as a GumStix or a linux powered microcontroller
• More testing of each component, and the system as a whole (unable to this time due to time contsraints)
• Calculate the power draw, rather than just assuming 4Ah would be enough

For an overview of the 2005 event, including all the final rules, see 2005 Rules.
For pictures of the bot, please see the Gallery

In the spirit of freedom, I've decided to release all of my code under the Creative Commons License which basically allows you to do whatever you want with my code, as long as you credit me somewhere. Enjoy!

Specifications
Computer	Mini-ITX, 800MHz, 512MB RAM, 20GB HD
OS	Windows 2000, fully updated
Motor Controller	$25 R/C servo-controlled unit
Servo Controller	KESEA-10 prototype (Serial->Servo controller)
Sensor Interface	KESEA-10 prototype (Analog/I2C->Serial)

Project Files
robomagellan.zip	137k	All of the files, zipped
robomagellan.vbp	917 bytes	Main project file, this is the software that runs on the bot
frmMain.frm	107k	[form for above]
functions.bas	5.6k	[utility functions for above]
rddf_record.vbp	836 bytes	Program to take the gps signal and create a cookie-crumb path of rddf readable waypoints.
rddf_record.frm	10k	[form for above]
sample_gps.log	33k	Sample GPS log
sample_kesea.log	18k	Sample Kesea output
quad_left.rddf	4.7k	RDDF for the left run, this had a narrow path with many obstacles
quad_middle.rddf	5.2k	RDDF for the center path, this would probably have been the fastest on-path route
quad_right.rddf	6.2k	RDDF for the right path, which was the widest and safest
FindCones.exe	119k	This compiled program uses a USB webcam to detect the orange cones. It outputs UDP packets back to the main program
behavior.vsd	67k	An overview (not the final code, but what we intended) of the bot hehavior