WiFi - Positioning

[PuddleJumper]

Greetings all.

Looking over some various positioning systems to have several robots all know there relative position to each other.  After a long deliberation and talks with various manufacturers, I have decided that WiFi Positioning Systems seem to fit my application the best.

My problem lies in that most of the companies on the market tie into a database that stores the lat/long of major WiFi sources.  I just want my bots to know where they are in relation to 2 things:

Where are they from me (and my handheld Wifi Device)?

Where are they in relation to the other bots?

The bots will obviously be moving so that complicates things.  My question is, can I set up a system where the bots kinda talk like this:
BOT0 says "I am X units in D direction from BOT1, I am Y units in D direction from BOT2, I am Z units in D direction from BOT3"
BOT1 says "I am X units in D direction from BOT2, I am Y units in D direction from BOT3, etc"
BOT2 says "I am X units in D direction from BOT1, I am Y units in D direction from BOT3, etc"
BOT3 says I am X units in D direction from BOT1, I am Y units in D direction from BOT2, etc"

Even if the bots are moving, given that I have a min of 4 dots, shouldn't I always be able to set up a relative position?  and if I wanted to have a standard place to be relative too, I could make BOT0 stationary.

Let me know if any of you have any insights to this?

Ben

[hgordon]

There's no way that I'm aware to directly measure distance between 2 WiFi radios.   If your robot's radio can measure signal strength, you could try testing with a directional antenna, and use that to triangulate from know radio locations.

[JesseWelling]

you might try googling quality of service triangulation. at least that's where I would start.

[Admin]

Whats your required position accuracy (acceptable error)?

Have you looked into using beacons?

[PuddleJumper]

I have looked into a  almost everything I could find, basically sucking google dry.

I have found everything from Cricket (A sonar, RFID technology) to Skyhook (A WiFi technology that is more wide-area based). 

I figured out with drawings last night if I use WiFi signals coming out of each bot, then I can measure the signal of each bot.  If I can measure the relationship between DISTANCE and SIGNAL I should be able to get an approximate distance (not sure of the accuracy).  Assuming WiFi is omnidirectional, this Distance will give me a radius of a circle that each bot is in compared to other bots.  If I have one BOT that holds still (A beacon) - I can say he lives at (0,0) from there I think I can determine the location of the other bots using some math, that I am sure I don't understand.

I was hoping for an easier way.  Admin - to answer your question - I would like to be accurate to within 3 feet (or a Meter).  If I need to put in a 5th or 6th bot to make it that accurate I can.

PJ

[hgordon]

You need to look for a different positioning technique - the idea you have is not going to work.  Radio signal strength is not linear, and radio waves don't take a single path between Point A and Point B.  If you want a nice briefing on the problem, take a look at the DARPA LANdroids project description - http://www.darpa.mil/ipto/solicitations/open/07-46_PIP.pdf

[PuddleJumper]

Ok.

If the idea for WiFi and signal strenght is out.  What happens when I start looking at technologies like this?

http://www.toysrus.com/product/index.jsp?productId=2318153  (Simple but it must work somehow)

and a higher tech solution

http://www.loc8tor.com (they say they can develop what I want for $100,000-$150,000)

Airman did you ever buy one of these and take it apart (the top example)?

Anythought on this?  Between this and the transmitter I am stuck.

[Admin]

To reword what hgordon said into a specific example, consider what would happen if there is an object between two robots - no longer a straight signal path.

Here is another idea, have each robot a different bright color, and put cameras on each robot. You can wear a red shirt. The size of each color (in pixels) would give relative distance away from each object/robot/person.

Oh, and why do the robots need to know where every robot is located at?

[Admin]

http://www.societyofrobots.com/robotforum/index.php?topic=1748.0

[hgordon]

If the idea for WiFi and signal strenght is out.  What happens when I start looking at technologies like this ?

Read my first response more closely.  I said that you can use a combination of directional antennas and signal strength to triangulate from known radio locations.  You won't establish position within 1 meter, but you still might extract some useful information.

[jabberwocky]

Why not try something like this Ant Farm?  The 'Ants' use IR to talk to each other, but i'm sure it would be possible to add wifi to this so that you can communicate with them all?

http://www.ai.mit.edu/projects/ants/

[PuddleJumper]

Admin, I read that link that you posted - that's how I found that toy thing and asked Airman if he ever took it apart and knew how it worked, I also ordered two to satisfy my curiosity on how they worked, range, etc.

and, the bots need to know where each other because I want them to swim, act, etc as a fleet without having to control all of them individually.  Picture a school of fish, they kinda swim around and do their own thing, but stay within a given range of eachother feeding a reef.


A directional antenna would be a great idea, but most that I have seen are big, clumsy and need to be located on the outside of my shell.   I originally likes the www.trollycar.com RFID locater - it's cheap, easy to use and is pretty darn accurate.  Put the field of view is only 64 degrees.   I am working with them to see if I can get distance to a tag in 360 degrees, but then use the 64 degree view to zero in if I need to.  Also the parameter of being inside of a shell is being discussed.

I guess the problem is 2 parts (much like the track human link):

#1.  A bot needs to know how far it is from it's LEAD BOT
#2.  If it gets to certain ranges or receives certain commands it needs to increase or decrease that distance.

I read the DARPA tech sheet and how signal strength works is much more obvious now.  I see why that would complicate things.

IR is out do to sunlight, line of site, small beam and the bounce nature of waves water and other stuff.  I just don't see it be able to fill the shoes I need filled.

Multiple colors is out, because each bot will look the same - back to the school of fish picture.

so far I love the Loc8tor because I know it can be done (or they say it can), but hate it cause they say it will cost $100,000 to develop.
I like the trollycar.com product because it is cheap and might work.
and I am excited to take a part the Toys R Us product because I can.

PJ

[Admin]

the bots need to know where each other because I want them to swim, act, etc as a fleet without having to control all of them individually.  Picture a school of fish, they kinda swim around and do their own thing, but stay within a given range of eachother feeding a reef.

ok so what you are looking at is basically swarm technology meaning:

• robots don't need to identify each other
• must follow each other
• distance apart must be minimalized
• a leader is required

This is what I would do. I'd color all the robots a bright red color (make the leader blue), and put a CMUcam on each. Id tell all the robots to move toward the color blue, but dont collide with anything. If a robot doesnt see blue, have it follow red. Then I'd tell one robot (the blue leader) to not chase color, and just go where I want the swarm to go. If some robots are following the middle mass of red (all robots in its peripheral view), and a few are following blue, then all robots will end up following the leader.

I simplified this for you to understand, but you get the idea . . .

[PuddleJumper]

Having moved away from WiFi - for cost, and other reasons... I have landed firmly in the RF box.

I received my package from Toys "R" Us, and was instantly reminded of how fun it was to be a kid again. 

I received the Agent Tracker from WILD PLANET.  It is basically (from my limited knowledge of things) a  TRANSMITTER and a RECEIVER.  You turn on the TRANSMITTER set it down and walk away.  Turn on the RECEIVER and move anywhere from 1" to 25ft away.  The RECEIVER beeps faster the closer you get (in a step progression) and there are 3 LEDS (BLUE, RED, YELLOW)... that indicate distance... SOLID RED is about 6 inches, and SLOW FLASHING BLUE is 25 ft.
 
http://www.amazon.com/Wild-Planet-Gear%C2%AE-Agent-Tracker/dp/B0007V62XQ

So,  Two part question:

#1.  How would I go about hacking this thing?  I basically want to strip everything off of it and just have the transmitter part of the beacon, and the receiving part of the receiver.  I would also like to tell at what point on the board the initial reading is coming through from the receiver antenna and how it is being measured.

#2.  If i wanted to try and start from scratch or make something that works on the same principle... is this the type of product I am looking at starting with?

http://www.sparkfun.com/commerce/product_info.php?products_id=153

Any information is helpful as always.

PJ
 

[Asellith]

I'm just going to throw some thoughts into this thread but I may be wrong. Can you determine just the direction each bot is from each other? Such as locate the different wifi signals directionally and get angles from a reference point and make the user send out a location signal as well. With all those angles you should be able to triangulate the position of each robot from where you are right?

The only problem I see without running the large amount of trig required to do this is that you may need to setup a wifi beacon at a position (for simplification parrallel to the user) a set distance from the user. Then with that known length and the angle of the user to the bot transmitted back to user and the angle from the bot to the beacon that should give you enough data to calculate the exact position of each bot then the users computer (which I assume is being used to control the swarm) will be able to feed the positional data to each bot. So all the bots do is send out the angles to the different sources and with the linear distance between each source known then the computer calculates position and sends it back to the bot. So the main number crunching is done by the user instead of the bots. It may be to hard to get those angles from a wifi signal so I may be way off the mark because I don't have any wifi experience yet.

Johnny B

[Admin]

Break it open and poke around with a multimeter until you find a signal thats readable by your microcontroller. Then just solder that pin to your circuit. I would suspect a good signal can be found near the LEDs or speaker (because its already processed output data).

That sparkfun product is a transceiver, so it doesn't do the same thing . . .

[airman00]

i was planning to hack it , but went with having a uniquely patterned belt which the robot would follow with vision.

the leds light up according to distance away.
so then have a microcontroller read which LED outputs are set to turn on and then the MCU will calculate the approx. distance.


For example

IF LED1= ON and LED2 = ON and LED3=OFF THEN
Distance = 5+10 + 0

Distance = around 15

you would need to find how much distance each LED accounts for.( aka experimenting)


Please post a video or more info about it if you hack the spy toy.

( i mite hack it, if i find a cheap deal on eBay)

,Eric

[Asellith]

I talked to one of my co workers a bit who is a former marine. He worked with a system called PLRS. (Position Location Reporting System) This system used multiple receivers and 3 stationary transmitters to determine position of a group of people/vehicles in a define area. This could work with wifi and would probobly work better then with the VHF radios they used decades ago for the PLRS system (pre GPS)

From what I understand from him the concept is that the "leader" or base station sends out a time stamped identifier. Then each follower in sequence recieves the signal and sends out the leaders signal and time stamp plus their time stamp and id. This repeats for all bots in the swarm and the more bots you have the more resolution. The math behind it is the tricky part. As I understand it you have each bot with several circles around it indicating the distances in a 360 degree arc from it to all other bots. Because with the time stamps you can using the speed of light in air calculate the transmittion distance between targets. Then you have position by plotting the circles and finding intersection points.

   An example would be the User at the base station sends out a string "01 11:00:00:00" Then the each bot receives and records this value. The leader tells unit 2 to report. Unit 2 sends "02 11:00:00:06; 01 11:00:00:00 11:00:00:02" This gives the leader the info it needs to determine distance from unit 01 to unit 02. This repeats with unit 3 sending "03 11:00:00:12; 02 11:00:00:06 11:00:00:07; 01 11:00:00:00 11:00:00:03" This is his time to unit 02 and unit 01. Then you have 3 distances: 01 to 02, 01 to 03, and 02 to 03. This gets exponentially crazy after a bit but after every bot has reported the cycle can repeat. With the high bit rates of a wifi connection you could get almost real time information with this setup. Obviously there is a limit to the amount of bots on each cycle based on the timed response needed for real time positional data. So if it takes 45 ms to complete a round with 6 bots and you want a faster update time you need to reduce the amount of bots or only refresh so many bots per cycle so some less important bots could be queued less to get more resolution from more important members of the swarm.
    The math is a little crazy so a base station controller may be needed or each bot could just keep track of the leader and 3 or 4 other bots to give it position and they just watch for their buddies or the ones closest to them at that time.

Wow didn't intend to get that long winded but I got excited about how cool this could be.

Johnny B