Results

We have had preliminary success with getting the robot to recognize the pendulum. Several problems have been overcome:

• At first, the camera's position meant that it could only see the pendulum when it was left of vertical. This was fixed by both moving the camera back (by disassembling part of its support structure) and by tilting the camera. Of course, now calibration becomes more of a problem, and it's less intuitive to look at our intermediary results, since straight up and down now looks crooked.
• We attached a black background to the robot to improve the vision system. There may still be problems with the camera auto-adjusting the brightness level, but we're not sure if this will cause a problem at this point.

Remaining: accurately and consistently detecting the angle, getting the robot to move, calibration, and putting everything together.