|
|
Post by Dingo on Aug 27, 2004 12:55:27 GMT 10
I've thought about this but can't find an easy answer. What is the best way for a robot to know where it is. For example if a robot had a map of a house's internals what would be the best way to have the robot plot it's own position in that map? A knownstart and then stepper motors would work only if there was a guarantee of no slippage (ie motors turn but no movement) A eletronic magnetic compass might work if there was some way of the robot being able to identify "landmarks" ie IR beacons etc. Therefore it could triangulate it's position. This would be expensive (in money terms) and would require heavy processing on the several inputs. GPS wouldn't be precise enough in the home. Is there some way of placing beacons or markers at known locations and having the robot read the distance to each marker and triangulate position. If so what technology would you use in the beacons? I.e. what could you read the distance using, IR? Ultrasonic? Laser? RF? something else? I know you can do direction finding for radio frequencies. In that case if a robot knew the direct from it to three markers it could calculate its position. Is direction finding easy enough to do on a small robot? Is it accurate enough? EDIT: this might be an interesting link www.nelnickrobotics.com/df_robot.html |
|
|
|
Post by bender on Aug 27, 2004 13:35:10 GMT 10
well that was the sight i was about to post.
i think that the stepper motor idear could work but dont take my advice becuase im not to good with all that kinda stuff.
|
|
|
|
Post by Dingo on Aug 27, 2004 15:38:55 GMT 10
At first I thought stepper motors would be the best and easist. Unfortunately all too often the stepper motor would be inaccurate due to lateral movement during turns, slippage on loose surfaces and less than expected movement on non-flat surfaces.
I think the RF direction finding would be a good idea. A little bit of trig later and you've got your pos. I think this is the most robust method to implement at least cost.
|
|
|
|
Post by Bones on Aug 27, 2004 17:05:18 GMT 10
I got a electronic compass on the way for mine. It uses SPI interface to give raw data for the position. www.pnicorp.com/productDetail?nodeId=c39bThe other way is a program that maps the walls and makes a 3D map of the world around it. But thats a bit more advanced. Using a ultrasonic wall follower maybe. Bones |
|
|
|
Post by ZapBrannigan on Aug 27, 2004 18:39:38 GMT 10
Without external markers.....
Slippage can be recovered with stepper motors, if you hit a wall. The robot could trace the length of the wall or obstacle and perform a best fit to its internal map maybe?
Alternately, using sonar, you can create a obstacle-distance map (prob need at least four directions covered) which can be compared to a known internal map to figure out position - lot of computation required though.
With 1 external beacon, triangulate; use rf so it penetrates walls. 3 Sensors on robot to measure time sensor receives beacon sig. Work out distance and direction to beaon from time that.
An alternative to an added beacon, some other emitting source? Three sound receivers that could listen for a TV?
|
|
|
|
Post by Dingo on Aug 27, 2004 18:59:38 GMT 10
I was thinking three beacons each with a difference signal and then you'd get through walls easily and use the known position of the beacons to plot the robot within a map. You could offload this computation to a PC by sending and receiving wireless data
|
|
|
|
Post by Bones on Aug 29, 2004 14:42:24 GMT 10
I spoke to a guy at work that has done a project like this and he said easy.
The robot he build was using a micro controller. It used a colour CCD camera which imported the info into the micro controller memory.
Then it would sift through this data and look for patterns like chairs etc. It would make a map of where it is with this info.
The project was to place some aluminium cans in some rooms and get the robot to search the rooms for 3 specific cans.
When found it would pick them up and return them to where it started. Here's the trick. Only red cans with a certain pattern on them.
Nice and easy. Yeh right. But that is how he said to go about it.
Other wise a sensor that reconises colours eg. walls, colours on skirting boards. Or put bar codes with invisible paint on the skirting boards that only A UV light can see for references. The robot can spin around with the UV light on and when the sensor sees one it can go to it and read the code. Thus knowing exactly where it is.
Bones
|
|
|
|
Post by Dingo on Aug 29, 2004 23:34:50 GMT 10
vision processing simple..........ummm.....  Yeah a CCD camera and enough computing power would be nice be at this stage cost prohibitive I've been searching for direction finding curcuits and/or "bug" detector curcuits and they are few and far between. There is heaps of stuff for "fox hunting" a game of orienteering using radio direction finders but these are designed for big environments. I'm just looking for one or two rooms - surely there would be an IC or something somewhere that does the hard work of Direction finding? |
|
|
|
Post by Bones on Sept 1, 2004 3:02:29 GMT 10
|
|
|
|
Post by Dingo on Sept 1, 2004 10:24:30 GMT 10
THANK YOU BONES!!!!
I've been searching for days and not finding too much applicable - finding way too many interesting these about radio but not applicable to positioning
These links are great!
BTW MY latest order from Oatley arrived (pretty damn quick this time) so I now have a two 433MHz transceivers to play with I'll post results as soon as I have time to set them up (way too busy right now though)
THANKS AGAIN FOR THE LINKS BONES - YOU ROCK!!!
|
|
|
|
Post by Bones on Sept 10, 2004 1:20:00 GMT 10
|
|
