I am very new to this group (joined today) and I have an idea for a consumer product that would involve creating a controller for a small outdoor vehicle that I would like to bounce off this group. The basic idea is to create a machine that will learn a path when manually controlled, the retrace the path automatically on demand. It would use a positioning sensor like the Hexamite Utrasonic system
to sample its position during the teach mode, then it would use the same sensor to compare it position to the original path. An accuracy of 2-4 inches would be needed over a 100' square area.
I have no experience in designing microcontrollers, but I can program and have a desire to learn.
Is this a feasable application? If it is, do any of you have any recommendations for controllers or kits to start with for a proof of concept? Have any of you used the hexamite sensors in your applications?
Ok - You got me! I was reluctant post anything base on automated lawn mowing. If you google this group with the string "lawn mower" you get about a zillion hits, so obviously it is everyone's favorite first automation attempt. Maybe it is the inventor in all of us who has a bit too much excess processor capacity during the manual lawn mowin' process (i.e. too much time to think!).
The idea is basically the same as one posted in '96 (do the google search) from a guy who wanted to use the xy postion data generated by a standard NC milling software package to generate a path for his lawn mower to follow. This made sense to me, being tied to a manufacturing facility that uses this software often. If a person could devise a machine that could record a a path around the perimeter of a lawn and permanent obstacles, you could use the technology similar to that used in computer aided manufacturing software to generate 'face-milling' paths within an area. A creative programmer could even allow for changing the orientation of the path to allow different mowing patterns.
This would be a simpler approach than something like the 'Lawn-Nibbler' project because you are basically making a machine follow a pre-determined path. You are not relying on texture recognition or intelligence to make decisions. You could do it more efficiently than the 'Robo-Mow' random mowers, because the path could be optimized much like the Machining paths are to reduce cycle time in the milling process. Looking at the Hexamite Postioning sensors and beacons, it looks like there is a system available that uses trilateration to sense current position. Using this sort of system, I would think a person could probably make this work.
Just a thought...I will still manually mow my lawn tomorrow! Jon
My great-uncle died a few years ago. His nephew gave the eulogy... they had worked together many times doing farming tasks (my great-uncle was a machinist, his nephew a dairy farmer)
The story started how Rudolf was a great guy, and a 'character'... Now, this was the '60's! The act brought lots of laughs, and much "familial fame". "... there was the day I drove to Rudy's, and as I drove down the driveway, I could see him sitting IN his living room, mowing his lawn!"
His trick... a cable-and-joystick control.
I think Hammacher-Schlemmer has a "simple" machine that Crawls(!) along the lawn... and slowly cuts it to size... Look... it's got about a week to do its thing... you: an hour? So.. as I said, look for low-amperage solutions that take time, and resemble a turtle criss-crossing your landscape.
Problem: when people find out what it does, it becomes subject to theft.
Joystick and a cable - no this guy was pretty inventive. These are the types of guys I love. You see this stuff in the 'Think it's New?' section of Popular Mechanics where they show stuff we think we recently invented only to find someone had the idea in 1910.
I agree, the low amp approach has merit. Either way it takes energy and time. My lawn is about 2 acres (I live in Iowa, everything is an acre!), so it takes a considerable amount of time to mow, fertilized and weed. If you could get it to follow a predetermined path versus the 'bounce' approach, you could devise nice accessories that weed and feed as well as mow. The little electric 'Robomow' spin-offs do the job, but they are only more efficient in the fact that the do not use fossil fuels. We manufacture equipment for the landscaping industry, so there would be many commercial applications for this type of application.
If it worked, you would reduce alot of the the inefficiencies involved in carting my considerable weight around my lawn. You could probably use a small electric machine like the Robomow product on a big lot because you are not wasting energy bouncing over previous paths.
My biggest obstacle is the lack of knowledge on this subject of robotics and controls. Which is pretty dangerous...especially when you consider applying it to rotating blades...
If you are thinking on lawn mower maybe this link could result interesting:
doesn't need fuel (is solar), and although it just follows a bouncing algorithm (with some improvement) I can say it's a pretty interesting invention. It can be helpfull the detection of the grass border as they do.
In my Masters Thesis I have studied the efficiency and robustness of some greedy algorithms to cover a region compared to a random bouncing path, and I can say that if you have some uncertainty in your heading direction (due to your actuators) a random bouncing pattern is pretty efficient.
And everything in Iowa is square, and somewhat level.
If you're serious on productizing this, start out with a platform which can navigate around your lawn without the dangerous whirling blades. You'll find this challenging enough with enough potential for damage. Make sure your fence is stronger than your motor. Park elsewhere. Most of your problems will show up in the mobility aspects.
I don't think you will really get much use out of the CNC milling path programs. I would expect there is less than a few percent overlap between what you want to do and what they do. In a CNC machine (or plotter or other such instrument) the environment is very controlled, and the repeatability of the tool movement very well known. An autonomous vehicle on a lawn surface shares none of these characteristics with a CNC machine.
I think that assumption is correct. If you look at the ultrasonic sensors you pointed at in your OP (openning post), you will see they have a 13 meter limit, of approximately 40 feet. An acre is 44000 sq ft, so you need a range of roughly 210 feet to cover a square acre.
I think it could be done sonically, but a (low cost) commercial system? I don't know of any.
So perhaps you might think about a larger positioning system. GPS jumps to mind. With the new differential systems, I've heard you can get something like a meter of accuracy (take this number with a grain of salt. I'm not up-to-date). Yet if your mower is a conventional 20,22,24" deck, ah.... you could miss some spots.
Well, there are systems already developed for aerial spraying systems which have better resolution. Perhaps you could use something from them, but I bet a meter accuracy does just fine for them too.
The short story is you have a navigation problem that falls in the cracks. Too large for room sized ultrasonic systems, too small for existing mobile systems. So if you tackle this you either need to find something close to your needs and modify it, or develop it yourself. In either case, this is 100x more difficult than the bit of the puzzle about making the runs overlap. And as well is the difficulty of the routines which keep such a vehicle in a "constant mode of 'course correction'" about 100x more difficult.
I am not trying to criticise, but to focus you on what you need to do to finish your problem. You might want to think about how tightly you want the mower's passes to overlap. If you know that, then figure how finely you have to be able to resolve a position on the lawn to be able to acheive that. Then you will know how fine a positioning system will have to resolve position at any point on the yard. I think you'll decide you need to be able to resolve less than half an inch (about a centimeter) reliably anywhere on the surface of the lawn. And that doesn't begin to address which point on the mower is at that point. So orientation of the mower become still a larger issue.
I would think that once the mower made a turn, all it would need to know is its next 'point' on the end of a linear mowing path. It could use odometry to maintain a heading with a 'ping' once in a while from a beacon to determine actual location and correct its heading. What is out there for RF beacons and sensor systems that would be available for a system like this? Are there any solor powered beacons out there?