Servo - July 2010

Position Controller

The most bare bones dead reckoning robots can get by with no sensors at all. This, however, usually entails a long period of trial and error because a sensor-less robot has no way of correcting itself and no way to maintain consistency in the face of less than perfectly consistent track conditions. The position controller, however, even when used as a simple encoder, can address many of those problems. Because the encoder can accurately calculate distance, and because that calculation can be done independent of the wheel speed, the robot’s trek through the maze will not be unspeakably compromised when the bot begins to slow down because of low batteries or if it hits a rough patch.

After measuring the distances in our maze and inputting them into our dead reckoning program (thank goodness for copy and paste), we were ready to test out the bot in our minefield of hot button issues. The program didn’t work perfectly the first time, and the Scribbler overshot its first turn and ran into a book on contracts. It was back to the code for us!

The trial and error of a dead reckoning program also presents a fantastic opportunity for young roboticists to learn the value of debugging. By announcing on a debugging screen how far it has gone, what maneuvers it has completed, or what speed it is travelling, users can deduce what has gone wrong and learn the invaluable technique of isolating the problem variable and fine-tuning it to solve a problem. For us, the problem was the distance, and after a couple of changes it was no longer overshooting the turn.

Even though our distances were getting better, another problem appeared. The Scribbler began to drift to the left. We checked to see if something had gone awry with our encoder mount, but it wasn’t interfering with the movement of the wheel. We deduced that the motors needed to be recalibrated. And while this might sound like a bit of an intimidating problem for novice tinkerers, the Scribbler documentation provides a fine explanation and step-by-step instructions for correcting the problem. A bit more trial and error found the correct calibration, and we quickly finished up the fine-tuning of the dead reckoning commands after that. In the end, our dead reckoning Scribbler was able to complete our maze in about 20 seconds.

Even with a successful program, dead reckoning still has advantages and disadvantages. We’d like to give a tip of the hat to dead reckoning’s ease and efficiency to program, but we give a wag of the finger to its inability to adapt to unexpected obstacles. If a wayward cat found its way onto the course, the run did not end well for either party.