Method for Robot to locate me
ChrisL8
Posts: 129
I wonder if I could integrate this with a robot so that it could find me:
https://www.thetileapp.com/
It is a sort of "tag" you put on keys and then you can find it with your cell phone. It sends out a signal so the phone can tell you when you are close to it and when you are getting closer.
Is anyone aware of anything already out there for localizing an object/place/person/etc. via some sort of RF signal?
The issue that comes to my mind here is that the robot is going to have to drive around to sample the signal to see which way to go. It might not be obvious that the signal is "that way", only that it is "strong here than there". That might require a lot of movement to take samples?
With the size of the Arlo though, maybe rotating in place with a detector at one edge would provide enough samples to determine a direction to go?
https://www.thetileapp.com/
It is a sort of "tag" you put on keys and then you can find it with your cell phone. It sends out a signal so the phone can tell you when you are close to it and when you are getting closer.
Is anyone aware of anything already out there for localizing an object/place/person/etc. via some sort of RF signal?
The issue that comes to my mind here is that the robot is going to have to drive around to sample the signal to see which way to go. It might not be obvious that the signal is "that way", only that it is "strong here than there". That might require a lot of movement to take samples?
With the size of the Arlo though, maybe rotating in place with a detector at one edge would provide enough samples to determine a direction to go?
Comments
With the Arlo, you could probably traverse the distance needed to take samples quickly. The hardest part would be developing an algorithm to figure out which way is the better way to go because if you have a long hallway and the signal emitter is perpendicular to where the Arlo is, and there is only a wall between the Arlo and the signal emitter, the robot should begin to move away from the emitter, then turn around to go back toward the emitter. This would cause the robot to not go anywhere.
I couldn't find much on how those tags work.
I think that to be located reliably a few systems would need to be used, or one sensor mesh system off the robot. There would be a lot of hunting. You can save motor battery by using a pan base and directional antenna if you're going to use a beacon.
I was thinking along the lines of what ValeT said. Could you script ROS to have the robot sample every room for an XBee RSSI?
If you have a low power chip antenna type XBee the RSSI will give you pretty good resolution from around 12' down to 3' away. From there out it would require something else like Erco's laser guidance system to bring it home.
It could be a stepping stone to a better solution.
Does ROS have any machine vision image recognition? That stuff usually requires good lighting and close proximity but should be easy to augment in.
http://www.parallax.com/product/27982
http://www.parallax.com/catalog/wireless/rfid
Just a couple of passing thoughts. But wouldn't it be great to make RFID tags you wear and your robot uses a dish type antenna to get your direction or just an RF Transceiver mounted on a servo that scans 0 -180 looking for the strongest average signal then turns in that direction. Dom..
Another way I was thinking about is using a thermal sensor like this - http://www.melexis.com/Infrared-Thermometer-Sensors/Infrared-Thermometer-Sensors/MLX90614-615.aspx
It's not cheap, but if you were to scan it back and forth you'd get a cross section of the heat. I've seen some youtube videos it picks up heat pretty far away. I don't think it would be hard to program in people recognition.
If there were more than one person in the room I wonder if it would be possible to block your body heat in a way that would be like a bar code. For instance your shirt has vertical stripes that the infrared sensor can pick out, it would know it was you. A lot depends on the conditions, and it's not an all inclusive solution. It could be thrown into the mix and help a lot though.
Yes, nice part. The MLX90614 was a Parallax product at one time but it's obsolete now. You can get eval boards from Melexis but they're a couple hundred.
https://solarbotics.com/product/35003/ It's a really cool part and I wish they still made sold it. It goes out 50 feet and has versions with spreads in the 5, 10, 20, 60, 90 degree so a 5 degree spread would get you really focused in on a small area. However RFID might be something else to look closely at.
When I looked at the first link I guessed they were using RFID in their little tiles and the other thing shown was a transmitter/receiver. The googled link I found had RFIDs detectable up to 1,500 feet which seemed a respectable distance for the task. By having an RFID "tile" in your pocket (no battery required) the robot could have a unique code in its tile detector. Unlike IR - (body temperature detection might confuse the robot) the RFID code is unique to you and that's what it would zero in on.
As I understand XBee it needs a unit and processor on both ends-- one end goes to the XBee Propeller serial pins and the other XBee probably plugs into a adapter board to supply power and converts serial to USB for your laptop. An RFID tag wouldn't need power. Only the robot would need power to rattle the tile and listen for a response. But the final approach might do well with a more local IR temperature human identification and maybe an ultrasonic pinger to get a rough idea of the object size and confirm with the IR sensor's direction.
Of course all this sensor hardware will have to feed the robot's current location, direction, obstacle parameters into mapping routines in order to back out of dead ends or to politely knock on closed doors. My dog follows me around if I have bacon in my pocket but doesn't knock yet. *
Go for it Chris....Dom..
*from "As Good As It Gets"
I think higher power will be the enemy. With the low power stuff you don't get a signal until you're close. With the high power stuff you get the same signal strength all over. These are just thoughts based on experiences with other stuff, I would try with RFID though.
How would you measure the RFID signal strength?
Thanks for the info on that sensor, I had no idea they were for sale at one time and discontinued. I have ordered one I think that with a Ping on it would be a lot of fun.