[HI Cap][MauiMakers] Laser Range Finders

[RogerI...@aol.com]

Does anyone have experience interfacing to a laser range finder? I mean the kind that actually does time-of-flight, multiple frequency phase shift, or interferometry in order to calculate the distance, as opposed to the much simpler (and limited distance) triangulation.

 

What I have in mind is setting up retro-reflectors at the corners of, or in various position around, a field (think the size of Kapiloani Park) and using a scanning laser range finder on a mobile robot to periodically detect the range to each of the reflectors and thereby calculate the robot's position.

 

My objective in this is to create a mowing robot that can navigate its way around a field filled with Christmas trees, cutting the grass between the trees. It would need to know its position quite accurately, to a couple of  inches, so as to not mow down the tree seedlings. During planting the farmer would use the scanning range finder to pinpoint the location of each and every tree in the field, so that  the mowing robot would have a map of the entire field.

 

- Roger Garrett

 

[David Cornejo]

You'd have to mount the optics above the height of the trees to get a
clear shot.

Have you considered computer vision? I think you'd need some sort of
a sensing system to detect transient obstacles (santa, elves,
presents, etc.) anyways and I'm sure it would be much cheaper than a
laser range finding system with sufficient accuracy

The commercial products use an embedded wire to define the borders of
the area and various sensors to detect obstacles (mostly pressure
sensitive) and they do a random walk within the confines.

You might also need to consider the regulatory & legal aspects of such
a system especially if this is for commercial use - lawn mowers are
very dangerous and a home built system without safeguards could cause
problems with the authorities and insurance types.

[Jerry Isdale]

Laser range finders are expensive (thousands ... http://www.robotshop.com/hokuyo-urg-04lx-laser-rangefinder-1.html)

I agree with David, a computer vision system would be easier, cheaper to implement.

You could paint or plant some fiducial markers around the area and use tracking software to determine the bot's location. 

example:  http://reactivision.sourceforge.net/

Lots of others.  Fiducials are commonly used in ARToolkits.

Basically they are simple (B&W/high contrast) images that can quickly be detected by computer vision software.

Once found in the image, it is a fairly simple matter of calculations to determine camera location.

Jerry Isdale

isd...@gmail.com

[RogerI...@aol.com]

 

 

In a message dated 5/13/2012 5:13:05 P.M. Hawaiian Standard Time, da...@dogwood.com writes:

You'd have to mount the optics above the height of the trees to get a
clear shot.

Absoulutely. The trees can get up to 8 feet tall, so there would have to be a mast on the mobile robot. 

Have you considered computer vision?  I think you'd need some sort of
a sensing system to detect transient obstacles (santa, elves,
presents, etc.) anyways and I'm sure it would be much cheaper than a

laser range finding system with sufficient accuracy.

Computer vision would most likely NOT work, unless it's really, really smart. The tree seedlings can be as short as three inches, while the weeds can grow much higher than that. Even a person using a walk-behind mower can have difficulty distinguishing the weeds from the seedlings.  

The commercial products use an embedded wire to define the borders of
the area and various sensors to detect obstacles (mostly pressure
sensitive) and they do a random walk within the confines.

That's for lawns that are mostly open space. These Christmas tree fields can be twenty, thirty acres or more with trees every six feet or so,of all different heights. Random walk would not work. It might be able to "bump-and-turn" for well-established trees, but it would mow right over the seedlings.

You might also need to consider the regulatory & legal aspects of such
a system especially if this is for commercial use - lawn mowers are
very dangerous and a home built system without safeguards could cause
problems with the authorities and insurance types.

Yes, that would have to be taken into consideration. At a very minimum it should have signs indicating autonomous mower in use, in addition to the standard no-trespassing signs.

 

[Jerry Isdale]

A laser system would cost tens of thousands to do well.

A mast on the bots would not fit under branches of the trees as they grow.

A vision system would use fiducial markers for position finding. It would not need to see the seedlings.

It would work similar to the laser but at a fraction of the cost.

At 20-30 acres, you would need them every 20-30yards... and have a fuel level sensor so the mower could know when to return for refueling.

Jerry Isdale

isd...@gmail.com

[Laurence A. Lee]

As my family owns a Christmas Tree farm (we sell in Salt Lake), I'd be very interested in seeing your project succeed.  I'd love to sic a robotic tractor/mower on the field and avoid the obstacles.

Please keep us updated if you make any headway, or if you'd like some additional help. :-D

[Jeremy Chan]

I agree, computer vision seems like a good solution. Fiducials can
also be used to determine 2D or 3D vectors for triangulation.
Combined with a magnetometer, you should be able to always know which
direction your robot is driving, and whether it is aligned to your
grid.

GPS could also help to cross check unique fiducial locations in case
someone decides to mess with your fiducials =)

Sounds like an awesome project that could be useful for a lot of other
things, like automated planting and harvesting.

Jeremy

[RogerI...@aol.com]

:

I agree, computer vision seems like a good solution.  Fiducials can
also be used to determine 2D or 3D vectors for triangulation.
Combined with a magnetometer, you should be able to always know which
direction your robot is driving, and whether it is aligned to your
grid.

The initial grid-ness of the field slowly disappears over time as individual trees are cut down (it's a cut-your-own kind of tree farm) and new ones are planted. An electronic compass is definitely needed in the robot.

GPS could also help to cross check unique fiducial locations in case
someone decides to mess with your fiducials =)

GPS only has an accuracy of ten-to-twenty feet, so it's only suitable for knowing "approximately" where the robot is within the field. It would be used on the robot, but not for the high accuracy needed to avoid the trees, which are about 6 - 8 feet apart.

Sounds like an awesome project that could be useful for a lot of other
things, like automated planting and harvesting.

I agree. I just wish there were someone (ideally here in Oahu) who would be interested in working with me on it.

 

- Roger

[Jerry Isdale]

Any system should have several independent position determination modes.

These are used at different times and under different circumstances, and also to keep check on each other.

For example a basic GPS & magnetometer (compass) gives fairly gross position/orientation. 

Accuracy is not good enough for the task (mowing while avoiding trees) but it does give a starting point and sanity check to back up the other techniques (eg fiducial vision orientation) 

A high quality GPS is good to much higher resolution than the 10-20ft, especially with differential, etc. but that adds significantly to costs.

The gps/compass would give initial location/orientation which could help the vision system have a clue where to look. As mentioned, it would also be good for sanity check in case someone moved (knocked down) the fiducials.

Wish I knew more about this and had time available to help.

Jerry Isdale

isd...@gmail.com