Balancing without the tail was difficult. It looks like this was due
to the motors subtly shaking the frame. The accelerometers are very
sensitive. A stiffer mounting for the computers might improve a future
iteration of the robot, perhaps with some rubber shock mounts to damp
vibrations. This is apparent because the angle
can be accurately
filtered out of the accelerometer reading when the motors are off.
An interesting measure to combat the drift of the gyro and balance
without the tail was to
use
gains that caused a lot of oscillation, so that one can get a good
estimate of the gyro bias.
Some teams did an excellent job balancing with tail, standing
stock still. Check out the videos.
The hardware was cruel. A wheel came loose, there were
deadly injuries, and the batteries kept dying.