robot navigation

[AllanBertelsen]

I have read some discussions concerning robot outdoor navigation. I want – as many other - to build an automatic lawn mower. I got an old laptop and picaxe experience. Problem is how the robot estimates its own position. GPS is not accurate enough. Visual landmarks are too complicated. The buried wire is not feasible. It has to follow a map.
Is it possible to use radio beacons combined with turning directional antenna and a digital compass?

 

[AllanBertelsen]

I searched for some techniques for this idea. The foxhunt people is using both 2m and 80m band for their RDF games. The 2m include large antenna, but the 80m sniffers looks just right for the job. If two "foxes" with a little difference in frequence are placed in the garden, it should be possible. Does anyone have experience with these devices? How accurate are they?
Here is an example:
**broken link removed**

 

[giaosucan]

may we use a webcam ??with webcam ,robot can see an object when it is moving :-??

 

[Papabravo]

RF can give direction. It cannot reliably measure distance on the scale of a backyard.

 

[akg]

may we use a webcam ??with webcam ,robot can see an object when it is moving :-??

LOS problem

 

[suicidalmonkey]

I don't have any experience with this technique, but I have heard it mentioned a few times and heard that it can be accurate.

Anyway, an inertial navigation system might do the trick.

I can't offer any more advice, but I'm sure you can do some research and make a decision.

 

[bloody-orc]

i'm thinking of 3 beacons.with those and with some(a lot) of maths you can position yourself quite accurately.

Also what i'm thinking of is a laser guided system. there is a laser pinter, that shows this robot, where to go exactly. only problem with this is trees etc. Tell me/us when this is an option (when you hav eno trees or anything). i'll explain further my solution.

 

[dknguyen]

Inertial navigation is nice but probably has too much drift. So far, the only for individual builders, the only feasible method of dead-reckoning is with map-like 3 IR/sonar beacons where the robot measures the arrival time difference of the IR and sonar signal from each beacon (transmitted simultaneously, but since the IR travels faster than the sonar, there is a time difference). THis is used to calculate how far the robot is from each beacon to form 3 ideally intersecting circles which will tell you where the robot is. More beacons can be used if line of sight is a problem. But the sonar doesn't work well outdoors with wind. THis is the most accurate DIY method available for indoors.

If GPS is not accurate enough, building your own local-area GPS with radio signals is the best way, but for individual users it's really hard to measure the signal travelling times because it is so small (not to mention lots of time synchornization errors between all the transmitters and receivers). A PICAXE is way too slow to handle anything like this, at all. You will NOT be able to pull off the required hardware.

As far as webcams, that's just really hard to process the data to track something (and then even harder to compare it to a map) unless it's an overhead webcam with disctinct colour differences (and the program is probably beyond the reach of a lot of people anyways). If visual landmarks are too hard for you, then probably so is any other method using a webcam. And it's also way too much data for a PICAXE to handle. You will NOT be able to pull off the required software. (well, maybe if it was an overhead camera indoors, but certainly not outdoors with an isosymmetric-view. Either way it would get overl complicated for what you want,

Don't try visual methods or radio-timing methods.

Inertial navigation might be your only method because accelerometers and gyros are fairly cheap and will help your robot stay on course. You might want to have some kind of beacon or something so the robot can regain it's bearings every so often because of the drift involve, perhaps combine inertial navigation with sonar/IR beacons so in the likely even that the robot loses contact with the beacons (wind disrupts sonar reception, IR line of sight interference, etc.). The robot can gauge it's position roughly until it regains contact with the beacons.

 

[AllanBertelsen]

Hi again
My garden is big. There are a lot of trees and other obstructions. So vision based methods, including IR will not be feasible. My idea is that the computer has an internal map of the lawns. By measuring travelled distance by traditional dead reckoning and compass direction, the robot can calculate a reasonable good position. It will know where to look after the 80m beacons. By getting an accurate direction of the beacons using RDF, it will be possible to calculate the position. So I don’t need to know the distance to the beacons – just the direction.

 

[AllanBertelsen]

This post is just to make things more clear.

A PICAXE is way too slow to handle anything like this, at all. You will NOT be able to pull off the required hardware.

I got an old thinkpad that is going to be a part of this project. The picaxes will be use to make interfacing circuits. It will also include wireless LAN and maybe a webcam. The webcam is just for sending pictures back to the house.
I found this circuit:
https://www.ardf.de/tbh/mf80cm.htm
It's an 80m minitransmitter. It is used for practising foxhunt as it does not reach as far. That could be my beacons.

 

[dknguyen]

Inertial navigation would be the thing for you then, but because the readings do drift, you need some kind of occasional checkpoints so the robot can reset it's bearings. An occasional IR pylon or something might do so that whenever a robot gets near one, it knows for certain where it is and then can continue on it's way, like a landmark.

An old thinkpad (I'm pretty sure most computers actually are too slow to perform radio timing). You need special dedicated high-speed circuitry for something like that. The thing about visual methods is that you need to have a better idea than just "using a webcam to see the robot." How exactly were you planning to pick out the robot from the image and then correlate it to a map using a webcam? (Presumabley from an angled downward looking view). It's hard enough to do this from an overhead view..

 

[AllanBertelsen]

Forget the webcam – its just for convenience. Then I can follow and enjoy the work the robot is doing from my armchair. It's not a part of the navigational apparatus.
If an 80m sniffer can be used for finding the direction of a beacon, then there is no demand for high-speed circuitry. You just have to rotate an antenna to find where the signal is best. I don’t want to us radio timing. Just to find the direction. RDF (Radio Direction Finding)

 

[dknguyen]

Yes it is. THe price you pay is that you have to mechanically rotate the antenna (or use an array of them). If you are able to find a directional antenna of the appropriate frequency (I haven't been able to), please share.

 

[bananasiong]

hi,
i'm currently doing a robot mower. it is not that high-tech, but it works!

https://www.youtube.com/watch?v=f24HH5MbG2U

 

[dknguyen]

WHere did you get a clear chassis like that?

 

[AllanBertelsen]

If you are able to find a directional antenna of the appropriate frequency (I haven't been able to), please share.

All contructions plans for 80m recivers for RDF include a directional antenna. Most use a ferite antenna,but loop antennas are also seen.

For example: **broken link removed**

 

[akg]

but would rdf serve ur purpose ?, since the accuracy will dependent on the rf/antenna beam width .

 

[freeskier89]

WHere did you get a clear chassis like that?

I wouldn't be surprised if it isnt just cut pieces of acrylic from a hardware store.

 

[bananasiong]

WHere did you get a clear chassis like that?

i bought the plastics and join them up using glue and screws. but that kind of material is not good for chassis.. not hard enough.. it looks very gay right?

 

[Paul Obrien]

bananasiong, does it actually mow anything?