3 replies to this topic

[Matt Taylor]

Hi All, This is my first post, So I'll introduce myself. I'm Matt, I'm a games software engineer, a tinkerer, and I just got a Netduino While I'm finding parts for a larger mobile robot, I've been looking at ways of implementing autonomous navigation around my garden. My ideal end goal is to have something like an automatic wheelbarrow that I can use to cart stuff around the place. I'm starting small - I have a small RC car, and driving that with the Netduino is my immediate goal. What I want to discuss here is using active beacons placed around my garden as reference points, and then using triliteration to calculate the position of the robot (robots?) from this. The area the robots would be active in is about the size of a football pitch, and is almost totally level, but has trees and hedges that I think rule out line of site based triangulation using retro-reflective beacons and a robot-board laser scanner. So I have a few thoughts, and a few questions. My first idea was using GPS, however, from using the GPS on my phone, satnav, etc, and then checking the wikipedia article, I think that he accuracy of this will not be fine enough. Ideally, I'm looking for something under 30cm resolution (the paths between my vedgedables are about 60cm). That said, I'm informed by a non-technical archaeologist friend that they use a GPS (in that they have a hand-held magic gizmo) system that is cm accurate. So my second thought was using local radio beacons and time of flight. However, the Netduino, and I expect nothing I can get at radioshack can deal with the minuscule time intervals that I'd be dealing with over the distance of 10-100m. I've heard of methods of using the phase difference between beacons of same the frequency to calculate TOF over smaller distances far more accurately, but I think I need to read up on the limitations of this approach further. One interesting way of fixing errors here is by having known location beacons that send out a signal, a known location base station that receives them, and sends out a phase-error signal, and the unknown location receiver that uses both of these to calculate it's distance from the beacons. I cant find the paper for that, as soon as I stumble across it I'll post a link. I've wondered if ultrasonic beacons would work. What is the range of a consumer ultrasound transducer in air? Would this cause problems with the local wild-life? I'm also dubious about using a sound-based aproach, as there's a cold-store near by that generates a lot of noise. The other biggie is Timing. With radio, and to a lesser extent ultrasound, the timing would be critical to the TOF calculation. Would a crystal oscillator be accurate enough to keep timing over a long time (say 1 month, or better a year)? Ultrasound could be synchronised with a radio signal though. So there's my current embryonic thoughts on this. Hopefully there's some people out there that are also interested in similar systems, and we can get some discussion going and hopefully find a Netduino based solution Thanks for reading, Matt EDIT: Here's some links: The NASA Robotics Paralax QuadRover is actually something very like what my long-term goal would be. Here;s the homepage: http://www.parallax....97/Default.aspx and a demonstration of waypoint navigation here: http://www.youtube.c...u/2/-d-u9ShZIxA Interesting to note that on the parallax product page, only GPS and accelerometers are mentioned, which would suggest that is's using just GPS and dead reckoning... hmm, that's food for thought. Actualy, on second thought, I may be misunderstanding this, and the NASA guys are using a beacon system on-top of the paralax rover. This has turned out to be an excellent source of info. http://www.doc.ic.ac...al/vol2/jmd/#rf The Motarola Mini-Ranger Falcon solves the short TOF by bouncing the signal back and forth a few times .. interesting. Sorry BTW for readers that this is a bit of a rambling stream of consciousness post. If anyone finds it useful, cool beans. http://www.doc.ic.ac...al/vol2/jmd/#rf

[antoniekruger]

Hi Matt, I've had a look at this a while ago....more the positioning with radio beacons. Here is what I had in mind. Looking at the GPS architecture - in its simplest form, the GPS position is achieved by trilateration. The receiver gets the sattelite ephimeres, matches it to a calender(orbit prediction) and runs a best fit of the sat data, checks it to set fix quality parameters and spits out the answer - lat, long and elev. Now, my point being - applying this to land based antennas(ground sats). So many variables are removed that it should be a simple calculation, assuming the land based antennas position is known. The solution would be in replicating the L1(phase) of the GPS system for the local antennas. This will increase the re-usability of many GPS libraries. As long as you set the elevation mask to very low. Like you said - rambling stream of.......ect.

[Coding Smackdown]

If you are planning to just keep the robot within your back yard I would think you could get away with just using a beacon style system. Ben Heck did this in one of his episodes out on element14.com where he used a couple of XBees and Ultrasonic Devices to create a Call and Ping configuration to allow his automated luggage to follow the user. The XBee on the luggage would send out a request and the target would then reply with a ultrasonic ping which would then let the Arduino calculate the distance and direction of the target. I'm pretty sure you could set up some type of similar system that would allow the robot to navigate your backyard. Since the Xbee has a short range you would probably need multiple stations to cover a good sized back yard. I'm interested to see what you come up with. Good Luck!

[drea]

If you're considering radio location you should check out Decca Navigation.

I'm not sure you could get the sort of accuracy you're after (and you might have fun with FCC), but it might spark some ideas.

Cheers,

Drea