GPS navigation

[sickrick]

Hi my name is Ryan...
I'm currently building a robotic lawn mower for my mechatronics TAFE project, and i want to use GPS or DGPS for boundary and obstacle avoidance along with other sensors (sonar and infa red  ect...).
Ive located what i think is a good GPS recever, its a picaxe connect GPS recever with aerial and should do the job... but my main concern is with the softwere and how to get it all to talk with each other E.G. in my mind i would picture having a separate GPS recever mounted on a post linked to softwere and you would walk around the property being mowed and mark out WAY or NAV   points to set up boundary's and key obstacle's to create softwere and little pictures of the robot and map and form there you could give the robot a set path to mow and a pattern with witch to mow...this may sound well and good in my head but i don't know were to go or what to do from here as Ive never done a thing with GPS but i cant wait to get into it.... also what boggles me is once the WAY or NAV points are established how to actually get the robot to steer ,accelerate,stop ect..when it reaches these points, but again "IN MY HEAD" i can Imagen having softwere liked up (wireless) talking to the robot  with a map of its set course and WAY points on the screen and the GPS will talk to the map softwere to establish were the robot is on the map...than the map softwere can talk to the robot and tell it were it is, were its been and were to go next... a bit like a CNC program.
i really have no idea and this all might be a load of rubbish so i would really appreciate it if you could help me out and do a tutorial on this....THANKS HEAPS....Ryan

[2Hektik]

Heya,

Sounds like a good project, can you simply measure the distance travelled instead, and wite some code in a microcontroller to go until you have counted enough turns of the wheels? GPS is perhaps a bit complex for fixed avoiding obstacles in a known environment as you will have to know where everything is anyways. GPS is better for tracking motion over non deterministic distances.

You say you have a GPS receiver... which one is it? What does it output? What the reciever has to say will determine how you write your code.

[Admin]

GPS in general has an accuracy of like 3 meters . . . how big is the yard you want to mow? I agree with 2Hektik that it probably is not a good idea for this. It works much better for huge areas of land/air/water (non-deterministic distances).

Measuring wheel rotations wont work cause you are operating under high friction non-idealistic terrrain (aka not a flat tile floor inside your house). Error will build up really fast.

You should consider using a video camera mounted somewhere. Basically have the camera mounted over head, and paint your robot bright red. Then the software just finds the x/y coordinates of the bright red spot in the camera view. This then corresponds to the x/y coordinates of your robot. This method is used for the robosoccer leagues. They even use different colors for each robot so the camera knows which robot it is looking at. Consider using the CMUcam2 (expect a lot of pain to get it to work right).

This is good in that you can have all your processing stuff done offboard and inside, while your robot dumbly just recieves wireless x/y coordinates => no different than what your GPS would give your robot, but much more accurate.

[sickrick]

HEY GUYES...
Thanks for your input...I haven't brought the GPS recever yet but the one I'm looking at is a add on for the picaxe microcontroller http://www.microzed.com.au/connect.htm, i work allot with the picaxe-28X and there fore I'm comforable wit it,also you can by a data logger that is just a plug in job for the picaxe. I thought this might help with th project.
 The yard I'm attempting to mow is 5 archers aprox 80m by 160m or so, its pretty big for a robotic mower, and it has a house smack bang in the middle.. is that to big for the camera idea?... but again i have not worked with that kind of vision before would you need to design your own softwere or can you by something to do the job
i thought it would be a good opportunity to experiment with GPS so i could gain more knowledge and it would be a step forward but i cant get any help from my teacher as he has no idea and cant be botherd(its a shame)...
i have read a little about DGPS witch is apparently accurate to a couple of inches. I don't know how true this is or how expensive either...
i did have another thought of using sonar beckons placed around the yard that would triangulate the robots position but when i started to think some more about it   it got a little crazy and i thought it would be to hard for me as i am only in my second year of studying robotics... but it would be cool to Be able to put something like that together...
I did think about using a encoder 2Hektik but there would just be too many things that could go wrong as Admin said the terrain would be just too rough for a start and i kinda wanted to try something new, but your right it would be allot simpler to program and prob make things essayer..
but if you guyes think it'll be to hard I'M UP FOR ANY SUGGESTIONS..... THANKS HEAPS... Ryan... CHEERS

[2Hektik]

Going back to the encoder for a sec... the idea pertains to rough terrain also. The error is cancelled by multiple in encoder the same way traction control is calculated for a car. But since that idea is out...

The sole reason GPS would be too inaccurate is because o' tha Yanks... There is a an error applied to the signal so that folks like you can't guide stuff around that well. It'd be all too easy for some religious zealot to wire up a cruise missile.

How will you stop the lawn mower from running into objects like the dog or a child? However you achieve this would be how i'd detect known edges.

On the idea of SONAR... Its really not that much harder than the GPS idea, however steer clear of the sound idea when you're not in water. Use infrared or a cam.

[sickrick]

HEY...i never thought of running multiple encoders  that sounds a little more interesting 
would you take a average of all the encoders or would you run it so that if one of the encoders  had a differant reading,  the one that has the odd reading is disregarded and then brought up to speed or reset...
 with the other sensors i am going to run  sonar and infa red,maybe some lower bump switches for safe measure.  I'm not shore about the camera vision but it sounds like fun I'm just not shore about the softwere side of things for the camera vision and how to actually see obstacles and convert them into program code and then do an appropriate action ... so maybe ill have to look at that side of things  THANKS HEAPS.. i appreciate it...Ryan

[Admin]

You should probably ask yourself what kind of 'coverage' you would like the lawnmower to acheive. I mean, is it ok for it to just randomly move around within your boundaries and hope that over time your entire yard will be evenly cut? Or do you need optimal where it does straight lines down your yard? Obviously the first is much easier . . .

Since your yard is very large, GPS will work. Probably better than the camera idea. At one point an error was added to civilian GPS, but I think it was like +/- 200ft or something like that. I believe, although I could be wrong, that this induced error has been reduced or removed entirely. After all, cars have gps now, so they must be accurate within at least 150ft (probably much more) . . . What accuracy does the gps device you will buy claim to have? That link didnt work . . .

I am still very much against encoders for position control. With no external reference, no matter what they will drift over time.

Consider using a digital compass to compliment your other sensors. They cost like $50 and you can get like 1 degree accuracy. Since it has an external reference, you can be 100% sure that your robot is pointed in the right direction.

And as 2Hektik said, you will also need a cat/dog/child avoidance system so you dont mow them over . . . perhaps sonar and/or bump switches, low tech, but easy to do.

[sickrick]

HI ... i would like the robot to mow in straight lines so i could have some wireless softwere managing its progress with little pictures , but it doesn't have to mow it all at once like over a perion of a week or so would be fine..
sorry that link didn't work   www.microzed.com.au    try this one and highlight  the modules tag at th top and then come down to the connect and in that folder you will find data loggers and the GPS unit CHEERS...
I think the unit has accuracy of 5m, if so this is no good...

[2Hektik]

try this,

vacuuming/mowing same stuff 

http://www.roombadevtools.com/

[dunk]

like has allready been stated, the standard GPS signal from the military satelites USED TO come with a broadcast error.
what this means that without a super secret US military decoding key all GPS signals USED TO tell your GPS receiver you are somewhere between 0 and 100meters from the real location.
fairly recently they stopped broadcasting the error because the DGPS network made it largely irrelevant.

DGPS (differential global positioning system) uses a simple trick. as well as receiving signals from your chosen satelites, you also receive a signal from the nearest ground based station that tells you how much the broadcast error is and in what direction.
this way using DGPS you could work out very accurately where you were despite the broadcast error.
obviously the nearer you were to one of the ground based stations the better.

for your project you could set up a similar idea. have 2 GPS receivers. a receiver in your bot and a receiver in the middle of your yard.
transmit to the bot the position of the center of your yard and it will be able to work out how much error is being caused by atmospheric interference and other factors. this should give you some very accurate results.

good luck and keep us posted.
(i have been itching for a project that lets me integrate GPS but my back yard is only about 5 meters square.)

dunk.

[BillD]

The US induced error code was removed by Clinton (May, 2000) during his presidency.  At this point GPS is accurate at about 6 meters 95% of the time.  With the WAAS (error correction) accuracy is down to 3 Meters.  You can increase your accuracy with the dual gps receiver system and basically transmit your own error correction.  You mount one at a know location compared to your program data and send the correction information to the robot.  Another way to improve the accuracy for a location is to start each session at a know location and do a simple shift of coordinates.  Error will increase over time so depending on the run time you may have to return and relocate from time to time.
    I too, am pondering the idea of making a gps autopilot robot.  Of course the gps tells you where you are at, not which way you are pointed.  You would also need to have a compasses to determine direction.           

[cooldog]

just a idea but could you map the yard with a computer and GPS then have a image on your computer and send it to the robot also could you use IR becon at each corner to avoid that spot then have the create a perimeter by liking all the IR becons?

i konw it is confusing