[Alan Anderson]

Our robot did something "entertaining" during an early match in St. Louis. We had just finished up our drivebase PID tuning, and had a simple "drive forward six inches" mode selected as a final test. Apparently, one of our wheel encoders developed a broken connection, so the software acted as if one side of the robot was nailed to the floor. What we saw on the field was this:

The 'bot deployed the arm, did at least two full turns in one direction while managing not to scrape the gripper against the field boundary, stopped, did it again in the other direction, ran diagonally across the field, barely missing both alliance partners and the rack, then stopped a few inches short of crashing into the field boundary on the other side of where it started -- not intentionally, but because autonomous mode was over.

"Drive forward six inches" had turned into "drive forward enough to measure 6 inches even with one side not moving, turn in place trying to correct the measured yaw error, overshoot because the robot is actually spinning twice as fast as the software thinks, turn in place the other direction trying to correct the overshoot, get the correction right this time because the too-fast turn is compensated by a too-fast correction, then take off at full speed trying to correct the accumulated position error." That was what I eventually decided was going on in the program, anyway.