Autonomous Positioning

[SteveO]

How does everyone plan to keep track of their position, especially in autonomous mode (but having several applications in user control as well)? Some ideas we've thrown around are:
The Hall Effect Sensors
Optical (Mouse?) Tracking
GPS (probably too expensive and hard to interface)
A small idler with encoder
The Camera

Use of a gyro would be necessary in all of these instances, except perhaps GPS (which could only be reliable within around 2 feet anyway).

I know a lot of these ideas are covered in other threads. I'm just wondering how your team plans to do it and what kind of success/accuracy you've had in the past with it. I don't see anyone being realistically able to cap the vision tetra in autonomous without some kind of positioning system.

[Tom Bottiglieri]

This was from a few years back, but still more advanced than what most teams have ever done.

http://www.wildstang.org/2003/video/...SOverview.html

[Max Lobovsky]

An accelerometer is not neccesary for all of those. You will see....

[russell]

there is absolutely no way gps would work. It would be easy to do, but a gps cant possibly get within two feet. Even with all of the new corrective systems a gps can only be accurate to about three meters which is like ten feet. And the competitions all take place in giant metal buildings. And even if you managed to get a perfect usable signal you would still only be able to tell generally which part of the field you are in. A radius of ten feet means that only if you were in the very center of the field would it necessarily show you as even being on the field at all.

[Tom Bottiglieri]

For some reason I dont think a GPS is legal on a FIRST field..

And even if it was, how would you get our tiny PICs to run that thing. Is there open source GPS drivers somewhere? Don't you need some kind of subscription to access the satelite?

[Ryan M.]

Our team will probably have a enoders on the drive wheels on both sides, and we may throw on a gyro just for the heck of it.

[russell]

you dont need any subscription to use a gps, it simply has a clock in it, then receives signals from the satelites that tell the location of the satelite, and the time of transmission. Using this they can compute their distance from the satelite, then when they get multiple signals they can use trig to figure their position. The only problem is that I am not sure it is legal to receive data from anything other than the radio modems. In addition to the problems I pointed out above of accuracy. And price. As for having a pic power one that shouldnt be a problem. Most GPSs have ports that they use to send their position and such to a computer or whatever. Lots of people use gpss on robots, just they usually dont want their position down to a foot or so.

[phrontist]

Max is right, accellerometers aren't required, or even desireable, as they are innaccurate. His plan is interesting, but very different from my own, and yet neither of ours use accellerometers or gyro's. I'll tell you a bit about my own.

Two optical sensors, manufactured by Agilent, that are rated for 800dpi at 14ips. Each optical sensor is in it's own housing with associated optics. One is mounted at the center of rotation, the other is mounted towards the "front" of the robot. Two axis of motion data from the center sensor and one axis from the "front" sensor in the form of quaderature pulses are read in via multiplexed binary counter chips. Estimated precision is a quarter of an inch. Estimated accuracy is less than a foot. Probably much less, but lets not get ahead of ourselves.

[SteveO]

Our idea is similar to what the StangPS does, we just can't come up with a very accurate way to keep track of our displacement. Optical sensors would work, but were trying to do something simpler due to our lack of team members and resources. We'll probably wind up using an idler with positive only motion attached to an encoder/hall effect sensor to help (but not eliminate) wheel slippage inaccuracies; using the gyro and displacement*cos or sin(gyro).

[Joe Johnson]

Originally Posted by SteveO

Our idea is similar to what the StangPS does, we just can't come up with a very accurate way to keep track of our displacement. Optical sensors would work, but were trying to do something simpler due to our lack of team members and resources. We'll probably wind up using an idler with positive only motion attached to an encoder/hall effect sensor to help (but not eliminate) wheel slippage inaccuracies; using the gyro and displacement*cos or sin(gyro).

If you don't hit things and you don't "peel rubber" on purpose, dead reckening is every accurrate and repeatable.

Jim Zondag from the Killer Bees (team #33) swears by his routine with keeps track of robot position, velocity and acelleration in both robot coordinates and absolute coordinates.

I have been begging him to do a seminar at Atlanta on the topic.

Anybody who knows him, please help beg for me.

Joe J.