Need help with positioning system ideas...

[brentonw2004]

Hello all. About a year ago I made a robotic lawn mower. Basically it is a normal lawn mower that uses a gasoline engine to spin the blade and spin an altenator to charge two 12V batteries. The batteries are then used to power a couple wheelchair motors, one attached to each wheel. I have a wireless camera attached to this and I drive it using a remote control. I know, I could be one of the laziest people alive. I would like to take this laziness one step further and completely automate the system. To do this I need to have a way for the robot to know its position in the yard. I do not want to use any type of buried wire type system. I have thought about two different ways of possibly doing this, but I would definitely appreciate ideas on which is better or even which would work. First idea is to use a GPS system. I do not know however much about the accuracy of these system. I have heard that most are only accurate down to 3-5m, this would be very unacceptable for my needs. However, I do not completely understand if that means that they could be inaccurate on the actual position, but how about accuracy on a local scale. For example, if I take a reading at one postion and move .5m away, will it recognize this change in distance accurately? It is not a problem if the system is not exactly accurate on finding the global position of the robot, it is however important that it accurately find the position of the robot on a local scale(one position relative to another local position.) I hope I explained myself good enough. My second idea is to use something like a radio beacon. I thought this could be achieved by syncronizing a clock on the robot with a clock on the beacon, then sending out a radio signal from the beacon to the robot with somewhat of a timestamp on it, then using this timelapse to calculate the distance from the beacon to the robot. I imagine that I would have to use at least three beacons to correctly calculate the correct position with no ambiguity. I need an accuracy of at least .1m. Would this either of these ideas be even remotely possible? I know that the beacon idea should work in theory, but would a typical microchip(pic/motorola) be able to work fast enough to measure such small time lapses? Any advice would be greatly appreciated.

 

[bryan1]

Hiya brentonw2004,
last decade I worked with a mate designing a ripping machine for farming that did 3 operations in one pass and our initial results were outstanding ( this is besides the point) However we both did notice an unmanned tractor going around seeding :shock: and on asking the farmer where we were testing our prototype he mentioned it was the latest in laser technology. Now that was over 10 years ago when lasers were the 'bees knees' but with laser's and ultrasonic's being so cheap I reckon thats the way to go with your automated lawn mower. Now if your making this up just to be a lazy bugger then I've given you a good hint to get off ya butt and do some research, afterall what else do you have to do with ya time. I'm not being nasty but after your initial question this the best I could come up with I reckon :twisted:

Cheers Bryan :evil:

 

[akg]

using GPS for ur purpose is not feasible , and u know the reason..having a radio becon as u explained will require high precesion and high speed electronics ckt..

now my idea would be similar to the working of the computer mouse. u will measure the x and y movements so that u can determine ur current location.
now the problem is u have to decide the starting location .. here a radio tx can help.

here is the scenerio for this type.
there is a radio/ultrasound/laser becon at a single location,the lawn mover will detect it and move towards it initialy.. now from this point as it moves for action it can use the 'mouse' principle to track its current location(ok now one shld move it manualy). after completing its task it searches for the becon and homes in to that .so that it is again at a 'fixed/predetermined' location

 

[brentonw2004]

Thanks for the replies so far. I have thought about the mouse type idea before where the robot would simply have a starting location and then count wheel movements to find its position. There are two problems with this for my design however. First, it programming it to a certain lawn would be a lot more difficult than using some type of radio signal/gps. The second is that in the event of one of the wheels slipping on say wet grass, then it would most likely lose track of its position and end up running around the neighborhood instead of the lawn.

Would either the GPS or the radio beacon ideas be anywhere close to plausible? Is GPS accurate on a local scale(one position relative to another close position?) If I cannot get this to work, then is it even possible to create a radio rangefinder? I understand that the radio waves will travel at the speed of light, so to measure how long it takes them to travel a short distance you would have to have a circuit running extremely fast. Does anyone have idea on how to construct such a circuit? How do laser rangefinders work? Any advice or even a decent link to some more information. I would prefer to use some type of RF signal to measure the distances if at all possible. Any info would be greatly appreciated.

Thanks!

 

[zachtheterrible]

I know this involves burying some wire but how about this: A bump system in which the lawnmower just goes back and forth across the yard, sort of like those automatic vacuum cleaners.

 

[akg]

is ur entire lawn can be seen from some place ?? . then a marker in the lawnmover + 2 cameras and a good digital image processing can track the mover ..how about that ?

 

[brentonw2004]

I appreciate the additional advice. I will definitely consider these ideas if I cannot come up with anything better or a way to make the others work. I know that it is possible to burry sense wire in the ground and bump into it, or to have some type of rangefinder system were line of site is possible. However I still prefer the idea of GPS or using RF beacons if possible. The reason for this is versatility. I would like to be able to easily program in other more complicated lawns. I could also see a lot of uses for such a system in farming. My idea of how this process should work is as follows. Set up radio beacon unit/units in a somewhat central location. Use a palm pilot/laptop to walk around the perimeter of the desired lawn, farm, etc. All this time the radio reciever would be plugged into the computer. The distances would be logged and used to generate a vector map of the lawn. This would then be downloaded to the robot and the radio reciever attached back to the robot. That is my idea of how I would like to get this to work. Thanks!

 

[mechie]

Global or Lawn Positioning System ?

If there is a clear line of sight accross the lawn then maybe two lasers could be used ? (see drawing)
Each laser is arranged to scan using a stepper motor, this allows a simple, known angle to be set and transmitted by the laser head - possibly modulated on the laser beam itself.
As there are two lasers at adjacent corners of the lawn it should be easy for the mower to calculate its position using basic trig.
If one laser is 'slaved' to the other - only starting to sweep and transmit after the 'master' has swept past it then there will be no problem for the mower deciphering which angle was transmitted by which laser.

Refer to sketch:- Imagine laser 2 (red) is the master, it sweeps anti-clockwise in 1 degree steps. Eventually it will reach an angle where laser 1 (blue) will see this and start its own (clockwise) sweep of the lawn. This should ensure that the mower will never see both lasers together.
If the lasers step in 1 degree increments then you can resolve a position to within 4 inches on a 20 foot square lawn; the mower must not move more than 6 inches without receiving another positional 'fix' via the lasers.

Dumb enough to work :?:

 

[akg]

I appreciate the additional advice. I will definitely consider these ideas if I cannot come up with anything better or a way to make the others work. I know that it is possible to burry sense wire in the ground and bump into it, or to have some type of rangefinder system were line of site is possible. However I still prefer the idea of GPS or using RF beacons if possible. The reason for this is versatility. I would like to be able to easily program in other more complicated lawns. I could also see a lot of uses for such a system in farming. My idea of how this process should work is as follows. Set up radio beacon unit/units in a somewhat central location. Use a palm pilot/laptop to walk around the perimeter of the desired lawn, farm, etc. All this time the radio reciever would be plugged into the computer. The distances would be logged and used to generate a vector map of the lawn. This would then be downloaded to the robot and the radio reciever attached back to the robot. That is my idea of how I would like to get this to work. Thanks!

i think that normal GPS is not usable in ur case .. u can try for the diffrential GPS for increased accuracy .

 

[Oznog]

I know this involves burying some wire but how about this: A bump system in which the lawnmower just goes back and forth across the yard, sort of like those automatic vacuum cleaners.

The Friendly Robotics Robomower uses a large loop of staked or buried wire. It's the same as the "Invisible Pet Fence"; it can detect whether it's inside or outside the loop as well as the moment it's on top of it. Inside the mower it has 4 large (like 1" dia 4" long) iron rods with a small drum inductor stuck on a PCB on top of each. One rod in each corner of the chassis. I'm not quite sure how this makes a pickup but that's what's inside.

They also use a front and rear bumper. This helps keep it off trees and pets. It also has a sensor that detects if the front castoring wheel drops down due to it being picked up and kills the blade and wheels if it happens.

It also has a compass sensor, otherwise it's very hard to keep a straight line on rough ground.

It's a "bump" system. This field of science is called "Painter's Algorithms", the goal here is to get the most efficient coverage when the software does not initially know the shape of the yard. Without the compass sensor and the ability to navigate a straight line, the strategy will fall through since it loses its understanding of its position and orientation in the yard so it's basically blind.

Differential GPS would be required to get the needed accuracy. There's a reference GPS sensor on a fixed stake and a similar sensor on the moving object and radio communication between them. The 30ft+ position error that comes from the satellites will be the same on both sensors so an accuracy of inches is possible. Sadly I don't think stock sensors can do this though, they don't even report lat-long to that degree of precision.

 

[mechie]

GPS Accuracy

I know that the Rockwell Jupiter is an old design but if this is the standard of these things ...

 

[zachtheterrible]

While driving home from work with my boss, he told me that there are GPS systems used in construction that can measure down to some rediculous figure like 1/16" I think it was. Not sure if this is true, but I'm sure it would be very expensive.

 

[Oznog]

While driving home from work with my boss, he told me that there are GPS systems used in construction that can measure down to some rediculous figure like 1/16" I think it was. Not sure if this is true, but I'm sure it would be very expensive.

Differential GPS can achieve remarkable accuracy with high precision equipment (I believe an extremely accurate reference clock is essential) and a low distance between the reference station and the measuring station.

The reference station can be part of the DGPS corrections network, or your own equipment placed in a fixed location (which will likely be much closer).
https://en.wikipedia.org/wiki/DGPS

 

[akg]

hello brentonw2004

so have u settled on some idea for sensing ??

 

[brentonw2004]

I haven't decided for sure what to do yet. I have been looking into differential GPS and I believe that it would definitely do the job, the only problem is finding a somewhat cheap model that I can use to build my interface. However, I have not ruled out the option of using radio signals. I would honestly rather make my own radio rangefinding circuit than use DGPS, but I haven't quite figured out how to make it work yet. My basic calculations for the radio rangefinder would require a chip working in at least the 2-3GHz range for my current idea to work.

 

[upand_at_them]

The Forerunner GPS by Garmin is for runners, so I imagine it would have to be pretty accurate. Especially since it calculates speed.

And returning to the mouse idea...Don't the new ones, without the roller balls, use IR to determine that it has moved? Can't something like this be employed?

Mike

 

[DigiTan]

If you had a least two GPS receivers and surveyed your site's location, couldn't you just send your own correction signals? Those embedded GPS modules go for around $15 nowadays.