# Wheel slippage w/ positioning systems

I’m working on a positioning system for our robot, and I would really like to account for wheel slippage as well. Here are some ideas I’ve come up with or heard from other people:

• Someone mentioned a non-driven wheel in another thread. I really liked that idea, but I think I’d have a hard time convincing my team we need another wheel tacked onto the robot, plus with uneven surfaces it could become a problem.
• An accelerometer. They are a little expensive and some people have mentioned noise, but I think if you see fast movement in your wheels (slippage) and very little movment (caused by noise) you could interpret that as slippage. If you are moving and slipping at the same time, the noise would affect your reading though. A 2-axis accelerometer could also detect side-movement if you are hit, so it might be beneficial to have one on there anyway.

By the way I’m working on a system with the gyro and optical encoders. With accelerometers this isn’t an issue, but other problems come into play.

So, are there any other ideas?

I’ve thought about this a bit, and on the plane back from Houston, this year, I thought that if you could monitor rpms at a given wheel, and also monitor the current going to the motors powering that wheel, you could compare the two values and experimentally determine whether your wheels are slipping. If you’re drawing just a few Amps (some arbitrary low amount) while you’re seeing high rpms, you might be able to safely assume that the given wheel is slipping.

Just something that might help.

Accelerometer is the way to go if you can get it to work. The beautiful thing about acclerometers is that if you’re slipping… you don’t move. It’s as simple as that. Your output voltage won’t change, and you won’t be tracked as having any movement.

Reduce noise with a little filter…

We’re in the process of doing this now. Digikey sells 3-axis accelerometers for \$18 I believe.

Matt

Accelerometers seem like the perfect solution in theory. However, to get from acceleration to position, you need to do a double integration. What this translates to is that even a slight amount of noise will be magnified by the integration and your positioning system will think it’s moving when it’s not. In our experience we encountered far less error due to wheel slip with our positioning system than we would have due to the double integration problem. Our robot would end up within a foot or so of our target location at the end of a 15 second autonomous round, and that was after doing complicated movements. Inertia-based systems have been known to drift by twice that amount or more when they’re sitting still on a table for 15 seconds.

I thought that if you could monitor rpms at a given wheel, and also monitor the current going to the motors powering that wheel, you could compare the two values and experimentally determine whether your wheels are slipping.

We actually had everything designed to do that this year but ran out of time to do the software. We had fabricated a PCB that implemented our current-sensing circuit from 2002 which attached to our 2003 positioning system. Would have been fun to try, but part of the reason we let it fall to the wayside was because our accuracy without it was much better than we had expected, making it unnecessary.