Robot Localization - Coordinate System

[MaxBot]

Hey all, I'm a programmer and want to start getting into robotics as a serious hobby. I've got a few ideas of projects I'd love to tackle but they require the robot to know exactly where it is.

This is a functional robot so I don't care how it gets the information.. nor am I trying to make it act human-like.

I'm thinking that there should be some way of doing this with some method like this:
- Place 3-4 objects in your yard that emit a Radio Frequency or other signal type.
- The bot has a sensor that can detect the direction that the radio frequency is coming from in each signal emiting object.
- From there it's extremely simple to construct a coordinate system. I do know how to do this already.


I'd go for GPS but I'm pretty sure that is not very accurate when it get's down to inches (correct me if I'm wrong).

So my main questions are
- Is there a pre-built system in place that I can purchase that will help me do these things?
- Is there any documentation that you know of that I can read for this type of localization?

[corrado33]

Just throwing out ideas here... but could you use signal strength?  The closer you are the stronger the signal will be.  (I have no idea how you'd do that.)

Or you could have different antennas on your bot and poll each one.  Like one on each corner.  Maybe one of the antennas will have a stronger signal and then you head in that direction.

Or have one antenna and make a sorta reflective radar dish and spin the dish around.  Keep checking the signal strength and when it's the strongest, head in the direction the dish is pointing.

I have no idea how you'd check signal strength... 

[MaxBot]

No, you don't need signal strengh to figure out your position.

I simply need some electronics consisting of 2 parts:
Part 1 (Emitter): Emits some signal (Placed in yard)
Part 2 (Detector): Detects the direction that the signal is coming from. (Placed on robot)

Then with three or so Sets of parts 1 and 2.. you can figure out your bot's current position.

Let's say you want accurate coorindates for a yard. You'd place three Part 1's (The Emitters) at 3 corners of your yard. If the robot can tell the direction that the singals are coming from then since you know the length between emiters you can figure out the exact position of your robot using a little trigonometry.

So really I'm just looking for Part1 and Part2. It seems like something like this should exist in the robot world - but since I am new I don't know what it'd be called or where I'd buy it.

[corrado33]

Ah I didn't think you'd be setting up a triangle.  That way it's easy to triangulate your position.  I can't help you with your emitter or detector. 

[MaxBot]

Actually I can see how a big array of detectors on the robot detecting signal strength could be the way to detect which direction a signal is coming from. This is probably how a direction determining detector device would work.

But I was hoping there'd be some type of pre-built electronic component that could do this instead of me having to hook up 360 or so detectors in a circle or something to deterimine signal strenth from every direction.

[dunk]

it's a far more complicated problem than it at first sounds.

you are correct that GPS will not give you the resolution you are looking for.

using ultrasound beacons and measuring time of flight would be possible but any studies i have read about using this technique have had fairly limited success.

radio beacons are not ideally suited to this sort of problem as RF signals tend to refract round objects and reflect off others making it very difficult to get accurate data on either direction or signal strength.

you need electromagnetic radiation with a shorter wavelength to eliminate the refraction problem....
light has a very short wavelength so only refracts a tiny amount.
it is also very easy for us to predict what it will reflect off. (avoid anything shiny near your light beacons.)

as for light sensors, either use 38kHz infrared sensors designed for TV remote applications and IR LEDs modulated at 38kHz for the beacons,
or if you are planning on having a lot of computing power on your robot, use a USB webcam as the sensor and regular LEDs as the beacons.

many commercial indoor robot solutions use IR light beacons in some form.


dunk.

[Billy]

as for light sensors, either use 38kHz infrared sensors designed for TV remote applications and IR LEDs modulated at 38kHz for the beacons,

dunk,
I've considered trying this but my yard is long and I'm having doubts about range. Any idea on what range can be made to work? Issues with sunlight? Other problems?

Do the sensors have a digital output or can you get them with an output that indicates signal strength?

Thanks for any help.

[dunk]

hi Billy,
yup, sunlight would reduce range considerably.
i'm not sure what sort of range you could expect in bright, sunny conditions... if i was to make a guess i would say 20 meters if you use lots of bright IR LEDs on the beacons.

on the plus side, the beacons are cheap so you could use lots of them, each emitting a different pattern.


the 38kHz IR sensors are digital.
you want to avoid anything light based using signal strength out doors as varying natural light conditions will make it impossible to calibrate.


dunk.

[garrettg84]

I've been looking into this myself. I actually thought about (and will likely do it...) putting a large quantity of neodymium magnets throughout my yard, closer together on the edges, and just use a hall effect sensor array on the bot. I am likely to get disk/rod magnets and use their poles to 'encode' locations throughout my yard. Almost like a large piece of magnetic media =) I plan on making the magnets readable using a compass to make sure I'm driving over them the right direction to get the 'bits' right.

I have yet to do any tests to figure out how deep I can bury them or what ground conditions can really screw it up, so right now, its just a theory.

[adanvasco]

Do you have a fence? How big is your yard? How many sides does it have?

If you have a wooden fence for example, you can use ultrasonic range finders and a compass to measure the distance to each wall knowing the bot's orientation. With the distance to each side, you can simply figure the relative position on the yard.

[garrettg84]

http://docs.google.com/viewer?a=v&q=cache:DhVUg06fxWUJ:www.insset.u-picardie.fr/file/Recherche/ISIE2010v17.pdf+detecting+buried+magnets&hl=en&gl=us&pid=bl&srcid=ADGEESiJoXuzAUeua96_AMSnD4FynBrBwe8CGb6ziaU4k-Th0QnLDT6ouIVyD4DMWc1Ws9KbvETGnZvPmYATZtMy8Nk6L3iJMS2LEWiaXbC-Z0onB9JaEjtjA_e9eKPArhbJb_i-jEel&sig=AHIEtbSV-z76dBwsx9idc9COua8_TpxA1Q

The above mess is a link to a pdf paper from a french university on using magnetics for localization. It is a bit more complicated using 3d compasses, but I think the idea is roughly the same as my previous statement about buried magnets.

[MaxBot]

Thanks a lot for the great info dunk and everyone.

As you said, it looks like this is one of those things that sound like it should be easy to accomplish at first, but reality has other ideas.

I was looking for something quick and simple to set up in the yard without installing tracks or magnets in the ground. I was also looking for a signal that would make life much easier by traveling through objects (houses, etc) like radio waves do. I figured if I threw down a few radio wave beacons emitting a signal and could detect the direction of each one then that would be a very simple solution for making a coordinate system and waypoints for a robot to follow accurately.

This would have been a tough and robust solution that could deal with many different terrains and the robot's shaky movement on uneven terrain. I am probably not interested in using IR or webcam since the bot would always need a visible line of sight to the beacons. I guess that might still be possible but but would obviously require many more beacons in my not-so-level yard.

advanaco, that's a good idea. I do have a fence but it's a split rail Also, was hoping for an easy, tough, and robust system that could be used in any yard without much fuss.

If i go forward with the project I'll prob either try many IR beacons or try to think of something outside of the box.

[adanvasco]

How about this http://www.societyofrobots.com/robotforum/index.php?topic=6194.0?

Check the links provided by Ro-Bot-X.