Request Robotic Lawn Mower

[ryan]

Hi everyone,


I'm planning to build an autonomous lawnmower. The robot should be less than 20 kg, its velocity should be 0.2 m/s, its running time should be some 30 minutes. Also the robot should have to bump sensors (one in front and one at the back) and it should move inside a perimeter wire (or buried wire fence). I have no clue how to build the buried wire fence.  Also I'll be using mulching blades to cut the grass. Can you please help me?

Thanks for your response

[pomprocker]

I think that if you run an algorithm like the roomba's do then you might need it to be able to run for a couple hours (the neighbors would hate listening to a lawnmower for two hours  )


build it as small as possible since it will be cheaper and easier to work with and its a prototype.

once you get that figured out then you could make a larger production model 

[ryan]

Thanks

[steferfootballdude13]

Your writing a dissertation and you want to make a robot that will cut your grass? Don't you have some research to do or something?

[Gertlex]

You might find it easier to use 3 sets of cutting blades in a line, diagonal to the direction of driving.  That's how at least some fo the commercial lawn-mower bots do it.

[ArcMan]

I'm also building a robotic lawnmower using R/C control.  Per your request for parts sources, I'm using Merits wheelchair motors from ebay, drive wheels and swivel casters from Harbor Freight, 33 AH SLA batteries and an old broken-down push mower.  If you'd like to see some very nice (R/C control) robotic lawnmowers in action, go to:

www.evatech.net

[Admin]

it should move inside a perimeter wire (or buried wire fence). I have no clue how to build the buried wire fence.

The basic concept is that the wire emits some signal at short range (maybe 2 feet?). Then when the robot comes close to the wire, it detects the signal, and runs the 'turn around and go another way' algorithm.

[Webbot]

Debugging software with rotating cutting blades could be a bit tense! Guess you've already figured out that your Number 1 piece of kit will be a remote cut-out switch and a manual panic button in case the remote fails. You don't want to accidentally give the dog a haircut.

It could be interesting to use something like Doppler Shift to detect things in front that are moving towards you (as well as just stationery objects). That way you could make the robot stop if someone/something is approaching it (which sort of gives you the remote cut out) - as opposed to just going around stationery things. Just a thought! Guess you've got enough on your plate.

If you are making a small proof of concept then perhaps you could use an R/C airplane propeller as the cutting blade?

[ryan]

Thanks you all for your suggestions. I have chosen the parallax motor mount kit for the drive system and the sabertooth motor controller.

For the cutter motor and blades I have not yet figured out which one to use?

Admin, how can I implement wavefront algorithm so that the robot can move from grid to grid and avoid obstacles (I have already read the wavefront algorithm tutorial) and then clear it's memory map and move again to mow other areas? For my first robot it is not that simple

[ArcMan]

Ah.  The Parallax motor mount kit and Sabertooth motor controller.  You've got good taste and money.  Perfect for the robotics hobby 

[steferfootballdude13]

For my first robot it is not that simple

easy, your first robot should probably be the $50  Robot. Its cheap and somewhat easy and gets you acclimatized to what you will be seeing. Making complicated robots early can be very bad. I Do understand your thought process though. Also if you make the $50 robot you can make a prototype of your lawnmower, to test what works and what doesn't.

[ryan]

I need to do this project ...so I have no other choice

[MadMax]

OK, then why won't you focus on creating something driving first. Then make it detect the perimeter it should be inside. If you've got that implemented, implement an algorithm to quickly cut the grass. THEN add the blades, you don't want it to act crazy because of a programming error with razorsharp spinning blades.

You also might consider a lot of safety measures, for example: you don't want the blades to spin if it picked up, if it bumps in to something etc.

[steferfootballdude13]

I need to do this project for my thesis...so I have no other choice

This is a somewhat off topic question but What is your Major?

[ryan]

Mechanical engineering.

I am trying to figure out how wavefront algorithm works. From Admin's tutorial it says that the robot has to have a goal cell after the robot has scan the area and move to that cell in the least time possible while at the same time avoiding obstacle.

In our case the robot has to move in each cell over the memory matrix and only avoid those that are considered as impassable. How to do this with wavefront Algorithm???

[Ro-Bot-X]

You are cutting grass, right? I believe you'll want it to look like a regular mower did it, gras cut going in 2 directions not haywire. So here are my thoughts:
Build a map with the boundaries and impassable cells marked. Also, mark a start cell and a goal cell (they may be the same point, but the robot will start going in one direction and arrive from the other direction). Have that stored in the memory (EEPROM) of the microcontroller.
Start going straight in the direction you prefer, then when the robot reaches the boundary, have it turn Left (or Right) move to the next row of cells, turn Left again, then go straight. If a impassable cell is reached, have the robot pass it and get back on course. You may want to use parts of the algorithm to determine which way to pass that cell.
Make sure you use encoders with PID controll and another mean of determining the position of the robot throughout the course (GPS, sonar triangulation to the surrounding objects). Have the cells a little smaller than the width of the cut grass line. This should work without major problems. Some tweaking may be required, depending on your lawn up and downs, since those will trick the encoders.

Have fun!

[Admin]

In our case the robot has to move in each cell over the memory matrix and only avoid those that are considered as impassable. How to do this with wavefront Algorithm???

Well, I wouldn't use encoders for this, because on grass in a backyard the slip error would build up insanely fast . . . but if you still wanted wavefront, your robot must declare a goal location as a part of the yard that hasn't been cut yet. After its cut, it should declare another uncut location as the new goal. But really, I'd just imitate the iRobot Roomba maximum coverage algorithm.

[ryan]

Thanks I have a case to argue for not using the wavefront algorithm. I'm not using camera vision also, i.e textural analysis.

The iRobot Roomba maximum coverage algorithm consists of what - errr moving randomly? Can you help plz?

[MadMax]

I would create an algorithm that would divide the garden area into squares. Then add a second algorithm that will make sure some areas will be removed of this square (because of trees/bushes etc.). Then you should align the robot with a side of one of these squares, and start cutting the grass systematic and efficient.

[Admin]

This should get you started:

http://www.google.com/search?q=roomba+coverage+algorithm

http://www.google.com/search?hl=en&q=robot+coverage+algorithm

[ryan]

Thanks 

For the sonar and IR sensors do I have to connect it directly to the PIC or should I use some kind of interface like the rangewizard?

For the perimeter wire does someone have a schematic? I have searched for pet electric fence schematic but did not got one.

[Admin]

For the sonar and IR sensors do I have to connect it directly to the PIC or should I use some kind of interface like the rangewizard?

directly to the PIC

For the perimeter wire does someone have a schematic?

this should help get your started:
http://www.societyofrobots.com/robotforum/index.php?topic=4733.msg39825#msg39825

[Terarius]

Theres another post like this here on the forum as ive seen. You might just check out youtube as well, many usefull videos there.
Weird site Doesnt seem that serious when a lot of cartoon figures jumps around on the pictures

[ryan]

Here's a schematic for the perimeter wire, is it ok? (moved below)
What sensor do I have to use to sense it?
Thanks

[colorclocks]

Ryan,  Sorry, but I can't read the pin numbers well enough to tell if it's ok.   The output section looks right.  Note that you want the perimeter wire to be the portion between the resistor and the transistor.  If you use the portion between the supply and the resistor, a short to ground could damage your supply.

I posted a schematic for the wire sensor in another robot forum.  I haven't figured out how to upload an image to this forum.  Until I do, here's a link to that post:
http://www.botmag.com/forum/showthread.php?t=896.  I tried linking the image directly, but it didn't work.

[ryan]

Colorclocks, I have re-uploaded the image. To view it more clearly right click and select view image.  For the sensor you are also using a 555 timer?

Also can the Parallax PING ))) Ultrasonic Range Sensor be used with the PIC16F877A?

[colorclocks]

Colorclocks, I have re-uploaded the image. To view it more clearly right click and select view image.  For the sensor you are also using a 555 timer?

That won't work.  Connect 2 and  6 together, but don't also connect them to ground.

Yes, in the sensor circuit, I'm using a 555 timer (in a monostable, or "retriggerable one-shot", mode) after the comparator.  But the comparator is the actual sensor; the 555 is just signal conditioning.  The 555 turns the square wave output by the comparator into a TTL level.

[MadMax]

Also can the Parallax PING ))) Ultrasonic Range Sensor be used with the PIC16F877A?

Those sensors can be used with any microcontroller that has a A/D converter (assuming they output analog values)

[ryan]

Is it ok now?
Also for the receiver part is the output analogue or digital?

How did you design your bumper?

[ryan]

Will a wire of 100 m be suitable for this circuit?