All right, I've got a direct-control drivetrain test work; the ROS node part works too. The next step is to write a general motion control cognitive process to start integrating motor control with sensor input.

For example, the motion control cognitive process needs to continually estimate its motion; if the motors are supposed to going and the robot isn't moving, the motors are stalled and will pull too much current - even at 3.3V, this will cause a lockup.

Here's a video: