Test runs went a little better last night. I went in a little earlier tonight and left earlier too. The drive in was a little spooky with the smoky haze from the fire near Boulder.
The good news? The robot made it around twice in a row! However, during the last several runs the robot was back to drifting into curbs. I'm still analyzing why and what I can do to improve the estimation.
Meanwhile, on the obstacle detection front...
|Propeller in 0.6mm pitch QFP, 8-SOIC EEPROM|
|Plenty of room for cute logos...|
The software to run the Propeller board uses a slightly tweaked i2cslave object (Tim Moore and AiChip Industries) from the Propeller forums. It also uses a slightly tweaked FullDuplexSerial.
One cog is dedicated to controlling the AVRcam, sending the colormap for tracking and then putting it into tracking mode and gathering bounding box data. Another cog makes the collected data available to the I2C master.
There's a diagnostic serial port on the board so I can see what the thing is up to. And of course an FTDI programming port for obvious reasons. The bi-color LED is used to indicate system status.
Eagle files: Schematic, Board
Source code (latest)