If I had a Rover and a square acre of land, and I wanted the rover to start at one corner, let’s say the bottom left, then travel to the bottom right corner, then move into the square one rover’s body width and go back to the left (picture mowing your lawn), until I finished at the top right hand corner, then send the rover back to the starting point…
What would be the best way to accomplish this? A buried wire? A radio signal? Program an Atom and use motors with encoders?
What would be the most reliable way to accomplish this?
Combine this: link with a good GPS module and it might work for you.
I do not mean to push Parallax products (I have a number of Jim’s creations and components), but Jim has not designed anything bigger than the A4WD1 and the Parallax Motor Mount and Wheel combo is about right for one acre.
I’m afraid I did not make myself clear with my shout out for help. I have a lot of motor/gear/wheel combinations that can traverse the acre.
I’m looking for the best way to control the directions of the Rover (motor/gear/wheel combos) so that I cover every square foot, I cover it in a predictable manner and I cover it in a repeatable manner. GPS may be the way to go, but maybe setting up an Atom to count the revolutions of a wheel to know that the Rover has travelled the length of the field, initiating a 180 degree turn, and travelled the length back the other way is correct. I see a lot of inherent failure with the counting wheel revolutions route, so maybe I should use electronic sensors to let the Rover know that it has passed a specific point and continue on till the next point, where it executes the 180 degree turn…
I can build a Rover to navigate the terrain. How do I control it and make it follow a path, cover the acre and repeat?
I did not mean to dismiss your advise about a good GPS unit. I know absolutely nothing about GPS and maybe that is a the way to go. Is GPS resolution so fine that it can discern the width of a lawnmower on an acre and make the Rover cover every square foot of an acre with accuracy?
Not much info given for the specifics, but I’d put an optical beacon on the bot with diretional sensors around the perimeter of the lot to determine the bot’s location.
GPS by it’s self won’t navigate you fine enough. You could improve it with other sensors, like gyros and accelerometers. And a good Kalman (sp) cfilter.
Counting turns and making exactly 180 degree turns won’t cut it either (pun intended).
I’d be inclined to set up some sort of beacon grid. Navigate back and forth between pairs of beacons.
The buried wire approach is probably a good one to consider also. With the Kalman filter, you can “add” the outputs of several navigation aids together to get a finer result then any individual sensor.
Hi, I would say for accuracy, the buried wire option sounds like the only method that would be fool-proof, but it would involve a lot of effort.
Any optical/radio beacon outdoors would not have the necessary accuracy of a foot or so within an acre.
I had been thinking about a similar issue and was thinking along the lines of a remotecamera on a high post that would have the whole field in its view and could track a colourful robot using roborealm and the path planning module. It would then direct the robot accordingly.
Not sure if its possible, but worth thinking about.
a burried wire is only going to get you perimeter detection. to get the type of accuracy you are looking for you will need to use multiple systems and combine the results to determine a position, or perhaps more accurately a position error that you use to adjust your navigation. a commercial band GPS unless it is expensive isn’t going to get you the kind of resolution you are asking for, I doubt it would even be useful information unless you manage to jump the perimeter wire. you might be able to use the gps results differentially to see how much you have moved but even then I think it would require a fairly expensive module. beacons, probably optical, set up at multiple locations are probably the least expensive options. you might even be able to power them using the same burried wire used for perimeter detection. you can possibly use a decent resolution black and white camera with an IR filter on it as a detector. combining odometry information with absolute perimeter detection with access to 3 or more beacons at any location should be able to resolve absolute position fairly accurately. what is the shape of your acre? the geometry of it will impact the number needed and power requirements for a beacon system. the type of terrain and amount of traction your rover has will also affect the accuracy of the odometry data as well. lastly, beyond designing a dog collar system some years ago, I’m not sure how accurate those systems are either as bury depth of the wire will vary and contribute to a position error as well.
Hi, I wasn’t talking about buried wire as a permimeter system.
The buried wire system we have here at work covers the enitre warehouse. there are runs up and down every aisle. The picker can load from any shelf on any aisle following the wires.
Buried wire is going to be the most accurate, but it would involve an insane amount of work and material.
Depends on how close an analogy the lawn mowing was, how big a rover you were intending (small or car sized),operating conditions (rain, dark, etc) and the accuracy you are looking for.
Camera system still sounds the cheapest and easiest to me.
There are some interresting snippets in the forums that could help you, although not on the same scale.
I feel that I have to disagree with not using GPS.
I have seen a video of a a small 6WD1 rover that, using an inexpensive GPS and three axis accelerometers, and sonic sensors for object advoidance, traveled 800 ft from a known starting point, turned around and came right back to the starting point.
I believe that sensor fusion using a Kalman filter, as suggested previously, might do the job inexpensively.
Out and back one time to where you started from in a relatively short period of time on a nice day is about the best situation you can get for that GPS demonstration. It is an example of using GPS in a differential approach. Without periodically returning to the reference point and re-establishing a home position with some other technique the changes in atmosphere alone over time will accumulate error in your position. Assuming we are talking about a lawn mower here, covering an acre with your 800’ example would require about 50 trips if the paths were about a foot wide. Maybe if you put a single beacon as your reference point in the middle of the short edge, and allowed the rover to re-home to the beacon a few times over the course of the 50 passes, it might work just fine.
With proper optics for the beacon seeker and a large IR source, a beacon seeker might work. Each beacon could be modulated with a different ID code and direct the mower to different areas of the acre in question.
ya I think you nailed the idea behind a couple of my comments on the head; the cost can get prohibitive if you are going to try and rely mostly on GPS but a less expensive module combined with other sensors can get decent results. The thing about the lawn mower robots in general is they don’t need to make regular and accurate patterns if they run frequently. There are entire product lines based upon the whole nibbler approach. Anyway we are still really just batting ambiguous ideas around here though and until wade actually tells us what it is he’s actually trying to do it can’t be much more specific than that. I think regardless there will be a degree of exerpimentation to find a good mix of cost vs. equipment to meet the goal.
I am simply trying to determine what infrastructure I need to build to allow a rover to roam an acre of land with precision and repeatability. The land can be flat or have rolling hills. Ideally it would be free of clutter and be mown on a regular basis. I am laying the groundwork for a rover that collects golf balls at a driving range or applies fertilizer to a field on a regular basis. Or it manuevers a particular piece of property at night for security purposes, but it is not driven remotely by an RC signal. It is following a methodical pre-determined route, it does it day after day, and it will complete the route the same way today that it did yesterday, and tomorrow, ad nauseum…
The next application might be 2/3rd’s of an acre and the next might be 40 acres. If you have a parking lot and you want a rover to sweep the lot every night of cigarette butts and stuff, without using a random algorithm that might somehow cover the entire lot (if you have enough time), how do you program a bot to cover that area in a pre-determined methodical manner?
I know about GPS and how to add in other sensors to make it accurate enough to do what you want. Is this a business venture? If you have a customer for this, then I can demonstrate a system to do what you want. Mike C
I do not have a customer as we speak, but I bet I can get one within a week or two. I am very interested in pursuing this. How can I contact you so I can discuss my idea.
I am getting a whole lot of very good advise and I think that yours is no exception. I’ve been reading up on all the information provided and there are multiple ways to do this. RFID tags can be used with alot of the different proposals to verify that the rover is indeed on the correct path when he should be.
