PDA

View Full Version : [Question(s)] Positioning devices



mattqatsi
09-01-2008, 05:28 AM
I apologize if this is the wrong forum to post it in, but I'm new here(and I intend to stay beyond my initial questions)... So please put it into the right place if I'm wrong...

What possible Positioning Devices are there for getting a robot's position with centimeter's accuracy... What possible methods can be used and what products can be bought...

Note: This is for the ION competition, if that makes any difference

4mem8
09-01-2008, 06:28 AM
Hi mattgatsi: Welcome to this forum. Is that the robotic lawn mowing competition? I was thinking NorthStar, But that is mainly for indoor navigation.

mattqatsi
09-01-2008, 11:17 AM
Hi mattgatsi: Welcome to this forum. Is that the robotic lawn mowing competition? I was thinking NorthStar, But that is mainly for indoor navigation.
Yupp, it is for the autonomous robotic lawnmower competition... I checked the NorthStar website, but can't find any products I could use, are there any other suggestions?

lnxfergy
09-01-2008, 11:18 AM
Depends what you mean by 'position'. GPS will give you the absolute postion, that is, where on the globe the robot is located, but only down to maybe 30 feet. If, on the other hand, you know where you are starting, encoders on wheels can tell you how far the wheels have moved, and from this you can extrapolate how far the robot has moved. Accuracy of encoders will primarily depend on the environment: slippery ground, carpet, etc can cause slippage of the wheels which results in inaccuracy. The farther the robot goes, the more these errors will add up.

Do you really need 'centimeters accuracy'? Or are you looking for centimeters resolution? In the latter, you would be recording how many centimeters you think the robot has moved, but over time the count may be off more than a centimeter. For instance, if the robot is going 100meters distance, asking to be within a centimeter is asking for 0.01% error. Thats probably just not gonna happen in the real world...

-Fergs

ooops
09-03-2008, 09:59 AM
First - welcome to the forum!

Seems that a laser beacon would help define where you are, the more beacons the more accurate. Sort of a local laser GPS system.
As for the wheel encoders, if "wheel slippage" is a concern, and accuracy is imperative, I would think about a separate set of "non driven" encoder wheels, they would just free wheel on the ground (I would hang them on some sort of sprung suspension) and would be less prone to the torque slippage that would inevitably happen with encoding the drive wheels. Of course you would gain an error if the ground was bumpy, from the additional distance traveled up and down by the encoder wheel. At the end of the day though it is still "friction based", and prone to at least some inaccuracy.

Is there a link for the mower competition???

Adrenalynn
09-03-2008, 11:28 AM
Some ideas related have been tossed around in this thread: http://forums.trossenrobotics.com/showthread.php?t=2205

The rules for S_Oly 2009 are out and my creativitiy is being redirected there now - back-to-back gold medals are a moral imperative for my team (;)), so I can't think about any other competitions for the next [cough] six months or so... :)

metaform3d
09-03-2008, 11:28 AM
I've always wondered if an optical mouse could be hacked for inertial guidance. Seems like with a change of focal length it could track XY motion over a relatively flat surface.

Adrenalynn
09-03-2008, 11:34 AM
Absolutely. I don't know if grass is going to help that little camera though.

As I understand them from having hacked on a few, the widdle high-res camera is imaging the surface looking for imperfections [as "landmarks"]. When you move the mouse, it's those little changes in position of imperfections in the surface that are encoding. I suppose changing its focal length it could probably track across grass. A scaled printout of a grass surface put underneath a mouse would probably tell you, right?

The same thing could be done in roborealm with a high-res camera pointed down though.

metaform3d
09-03-2008, 02:57 PM
A scaled printout of a grass surface put underneath a mouse would probably tell you, right?Grass has a 3D structure, so it's going to depend on whether the imaging system could cope with changes of parallax.

I think that a camera pointed down would make an excellent sensor for a lawn mower, both for measuring position and for detecting the edges of the lawn area.

Adrenalynn
09-03-2008, 03:18 PM
Even though it has 3D structure, it's only looking for microscopic pixel differences, right? I guess if the grass under the "mouse" were moving in some counter-relation to the mouse's camera, that'd suck. Like dragging a piece of paper under an optical whilst moving the mouse in the opposite direction [that's a tough experiment to conduct, but with some gnashing of teeth, I tried it and it does screw things up...]

darkback2
09-03-2008, 03:30 PM
I've been tossing this idea around for a vision based outdoor robot. How about you have roborealm center a camera on an object of unusual color or shape or something...and then measure the angle of the camera servo. If you do that with two object, then you can calculate the intersect of the two angles and you have your location...This gets messed up if you have a blindspot...but in chapter six task two I sort of play with the idea of a lawnmowing robot that doesn't know where it is...but also doesn't care. If it stays within the boundary wire, and moves at a constant angle in relation to the yard (compass) then it should be able to mow the entire lawn. Have it do that twice from two angles and you are doubly sure.

DB

robot maker
09-03-2008, 03:40 PM
if you need a outdoor positioning system,most like you need a system that combinds all type of sensors ,because the ground changes alot,gps to get it near the area,then camera looking down or xy camera under the robot to scan the area,sonars for rocks and other stuff that gets in the way,wheel encoders and compass for direction ,speed and distance travel,maybe laser beacons,but all will have errors and problems ,but adding all together ,will cut down on error alot
for camera will need a fast processor or fast computer