Best way to make a tracking device?

rwgast_logicdesign · 2012-11-11 19:54

So ive been working on a bot, I also have 4 dogs. My dogs are nuts and I im not sure how there going to react to a robot rolling around. I know the one in my pic has alot of fun with RC cars! Also my bot is going to be shooting lasers all over the place, not good for an animals eyes.

One of the first things I need to do is come up with a system where the bot can figure out where the dogs are so it can firstly turn its lasers off, secondly take action to interact with each dogs personality, i.e roll away if the dog is after it. Whatever I use for a beacon needs to be small enough and workable on a dogs collar. I had thought about pulsing IR beacon at different frequency's, so the bot would know if a dog around, what dog it was, and what direction the dog was relative to the bot. But I thought about this and there are some things id like to be able to pull off where an IR beacon may not be the best idea, like know where the dog is even if not in the same room. I also want to be able to follow behind a dog about a foot. In reality I really need to be able to track the dogs location when its within 50ft or so. GPS seems ideal except 4xGPS=Too much and the qccuracy of gps isnt to great. Im thinking theres got to be a good way to do this. If need be maybe I could Send a BT signal from the dog with its x,y position but im not sure how to get its posisition or use some universal x,y system the bot can locate.

Mike Green · 2012-11-11 20:41

There's no simple way to do this. As you've noted, GPS isn't accurate enough. IR is line of sight. Dead reckoning isn't easy or accurate over the distances you're talking about (for the dogs' collars). I'd favor an IR beacon on the dogs' collars with a different frequency for each dog. Here's a detector for the IR modulation. You'd need to identify the frequency involved in software as well as measuring the signal amplitude so you can look for peaks as you scan the environment.

Circuitsoft · 2012-11-11 20:45

GPS can be pretty accurate, and there are cheap ones available.

http://dx.com/p/et-662-gps-engine-board-module-with-sirf-star-iii-chipset-80038

http://dx.com/p/gps-external-digital-antenna-with-mmcx-connector-and-3-meter-cable-1-5ghz-14926

http://dx.com/p/em-411-gps-engine-board-module-with-sirf-star-iii-chipset-80037

https://www.sparkfun.com/products/10919

rwgast_logicdesign · 2012-11-11 23:24

idk cant gps have up to 30meters of error? i need about 6inches of error.

if i go with ir could i some how tell the distance from the dog. i mean i could ping that genral direction but what if i ping through its legs?

i dont even know how to go about this but i thought maybe i could make a coordinate system based on a map of the house and then transmit the dogs coords

teganburns · 2012-11-12 01:36

Yeah a gps wouldn't do for 6 feet let alone 6 inches...

Mike Green · 2012-11-12 06:16

Maybe you could use a laser rangefinder once you locate the approximate direction of a particular object (dog collar). Frankly, you're unlikely to be able to do what you want cheaply or easily. This is a major problem in robotics. Often it's solved using multiple fixed video cameras with a wide view of a whole room or warehouse space and some very very sophisticated object recognition or multiple beacons on the objects and tracking software that locates objects in the space and triangulates them based on the known locations of the cameras.

Maybe you could use two CMUcam4s a short distance apart on a short mast so they have good overlapping views. You'd have a slowly blinking IR beacon on each collar. The two cameras would capture the position of the beacon and you could compute the angles and thus the approximate position and distance to the collar. A separate IR sensor could be used to time the width of the IR pulse for identification.

Duane Degn · 2012-11-12 06:36

Mike Green wrote: »

Maybe you could use two CMUcam4s a short distance apart on a short mast so they have good overlapping views. You'd have a slowly blinking IR beacon on each collar. The two cameras would capture the position of the beacon and you could compute the angles and thus the approximate position and distance to the collar. A separate IR sensor could be used to time the width of the IR pulse for identification.

Or two Wii Mote cameras? (Less expensive than two CMUcam4s.)

You might be able to do it with one Wii camera and use three IR LEDs (not colinear) on the dog. I think this method (a single Wii camera) might be harder to get to work since it's unlikely you could get the three LEDs spaced far enough apart from each other to get much precision in the measurements.

xanadu · 2012-11-12 09:00

Sounds like you need a proximity device. You can use low powered XBee and the RSSI pin, works great!

http://forums.parallax.com/showthread.php?139684-XBee-Locator-Beacon

You will need to incorporate a directional antenna on the bot, shouldn't be too hard.

And not to be a nag, but using visible lasers outside can get you in a heap of trouble these days.

localroger · 2012-11-12 09:42

Set up some synchronized ultrasonic beacons and use time of flight to estimate distances and triangulate.

Circuitsoft · 2012-11-12 09:57

Ultrasonic beacons work. Another idea, from a coworker, is to put a cheap, EOL'd bluetooth headset on each dog, pair it with a bluetooth adapter on your robot, and use RSSI to determine direction.

rwgast_logicdesign · 2012-11-12 12:22

well i had thought about ultrasonic, something like taking the recive side of the sensor on the dog, then when the bot recives the transmit from the dog u have distane. the problem is how in the heck could you posistion an ultrasonic sensor on a dog so the bots sensor could interact..

the bluetooth idea sounds intresting would 5 cheap ebay modules 1 master and 4 slaves work, instead of headsets? what is rssi

Circuitsoft · 2012-11-12 15:22

lmgtfy.com/?q=RSSI

xanadu · 2012-11-12 16:00

There's a way you can do it with no electronics on the dog.

Put a camera on your bot, download roborealm. Put something on the dog that could be a visual cue, or take a bunch of pics of the dogs from multiple angles. Roborealm will pick them up. The camera will always need to be scanning around you though.

You might need to use an wifi IP cam to send the video to a computer then commands back to the bot via xbee, unless your robot has some decent CPU power.

Some of the projects you mention are pretty intensive, you'll find a bunch of different solutions but the best way is to implement a few of them then put an arbitrator in the mix for accuracy.
Also my dog loves it when EMIC talks to him.

localroger · 2012-11-12 16:57

If you want to use RF you might want to look at these products:

http://www.rentron.com/PicBasic/RemoteControl.htm

The low-power transmitters and receivers are under $10 and the receivers have RSSI output which you could probably read with a delta-sigma ADC on the propeller. RSSI is a lot more approximate than ultrasonic time of flight but it has the advantage of going through dogs. You could drive those things with an extremely low-power and simple circuit (think 555 sending a different width pulse for each dog, at slightly different intervals so that if there's interference it will self correct). If you just want the bot to shut down when a dog is too close it could be as simple as looking for too-high RSSI and shutting down (you might not even need ADC, just a comparator).

rwgast_logicdesign · 2012-11-12 18:05

Hmmm ok I will look more into rssi either using bluetooth or the RF modules above. Tracking via rf is exactly what i wanted to do but figured it would be insanely hard. As far as the camera solution goes, im just not ready for machine vision yet. I would like to have my computer wirelessly recive a camera stream from the bot to send commands back but im really looking for the cheapest solution that will work, ive been told the ebay cams drop alot.

As far as talking to the dogs I want to do this too and im hoping to have the bot have voice recognition. Originally I was going to use an enmic for speech but the more I think about it I think using a Rasberry PI is going to be the most bang for the buck as I can use some easy test to speech and voice recognition software all for less than an enmic, the PI would also be suitable for openCV or some machine vision software once they release the camera module

Duane Degn · 2012-11-12 19:45

localroger wrote: »

If you want to use RF you might want to look at these products:

http://www.rentron.com/PicBasic/RemoteControl.htm

The low-power transmitters and receivers are under $10 and the receivers have RSSI output which you could probably read with a delta-sigma ADC on the propeller. RSSI is a lot more approximate than ultrasonic time of flight but it has the advantage of going through dogs. You could drive those things with an extremely low-power and simple circuit (think 555 sending a different width pulse for each dog, at slightly different intervals so that if there's interference it will self correct). If you just want the bot to shut down when a dog is too close it could be as simple as looking for too-high RSSI and shutting down (you might not even need ADC, just a comparator).

There are also lots of Nordic nRF24L01+ transceivers available on eBay for less than $3 each. These Nordic modules also have a RSSI output. You can get about 10 Nordic modules for the price of one XBee. The Nordic have lots of cool features (2Mbit/sec transmission, error check, auto retransmission, etc.) but they do require more pins than a XBee to control and they're not as easy to control as XBees. But for less the $3 each (I think you can get them for $2.50 if you buy ten at a time) they're hard to bet.

SparkFun also makes a little Nordic fob with these chips.

rwgast_logicdesign · 2012-11-12 19:54

What about RFID would it allow for rssi, seems like cheap rfid tags would be the best to actually put on the dogs.

xanadu · 2012-11-12 20:12

rwgast_logicdesign wrote: »

What about RFID would it allow for rssi, seems like cheap rfid tags would be the best to actually put on the dogs.

RSSI works depending on the type of signal. If you read up on how RFID works you'll see why some can and some can't. This is so far beyond my knowledge of RFID I don't even know why I'm still typing lol. If you read up on RFID you'll find a ton of info. But the way most RFID works like anti-theft at a store it would be very hard (maybe impossible) to return a signal strength.

Best way to make a tracking device?

Comments