Answers about auto-balancing

[archiver]

Posted by Andy Baker at 03/13/2001 10:56 AM EST


Engineer on team #45, TechnoKats, from Kokomo High School and Delphi Automotive Systems.



Many teams have been asking us about auto-balancing... and I'm thinking that we're going to get more questions. So, I'll try to describe what we do.

First and foremost, it is not just software. The software routine this routine depends on a VERY high traction system, and a powerful machine. We are fortunate to have both. If you have alot of traction in your system, but not much power, your motors will not be able to push the robot quickly enough to get the reaction required to counter-act the bridge tip. If you have power, but not traction, then you'll spin your wheels and not get the mechanical reaction needed.

Here is the process we follow to balance:
1. The robot travels up the ramp in low gear (high torque), pushing one goal and pulling the other. The driver is driving in a controlled motion at this time.
2. Once the robot gets with a foot of being 1/2 way up the ramp, our operator pushes the "Auto-Balance Button".
3. Once the button is pushed, the robot takes over. It now travels very slowly up the ramp... while the gyro is looking for an angular change.
4. Then, the bridge starts to tip, and the Gyro sends a signal to the controller, sensing this angular change.
5. The controller then tells the drive motors (2 Bosch and 2 Fisher-Price) to back up during a timing loop (for 10"-12&quot... really fast.
6. The controller then switches the direction of the drive motors to drive forward about 2"-3".
7. Voila! A balanced robot with two goals. There is enough of a sweet spot that the amount of balls in the goals does not matter.

The second direction change is hard to see... but it stops the robot's momentum and keeps it from backdriving down the ramp.

There is one big problem with the Gyro... noise. We filter this noise two ways: by mounting it to SPI dampening sheet (page 151) and by a software filter.

The fact that TechnoKart can drive itself and two loaded goals backward once the bridge starts tipping is a credit to it's drive system... not so much the software. After seeing other robots try this, we find ourselves lucky to have the right combination of power and traction. I think that this is the real trick. The software can tell the robot to back up for 500 milliseconds, but if the motors do not have enough uuummmph or the wheels spin, then this program may not be any good.

Now, keep in mind that I'm a mechanical designer, not a software guy. According to our programmers, the code is not that hard... the gear switching system is tougher, and so are preset positions being fed from potentiometers. The thing about this gyro code is that it works elegantly.

So, if you want to do this... get your traction in place and make sure you have enough power. Don't just tweak the code.

Andy B.