Re: [mowbot] RFD: Define reference 'grassbox' for mowbot

byron nospam at
Tue May 12 20:54:08 EDT 1998

> At 04:04 PM 12-05-98 -0400, you wrote:
> >I've been on the list for three years. It has had some interesting
> information.
> >
> >However after all this time I still feel like I'm lacking to knowhow, skills,
> >and equipment to develop a working mowbot.
> >
> >Early on in the list development, there were raging arguments about how
> >to develop such a beast. Personally I think that was much more informative
> >and productive than what followed, very few people discussing their own
> >personal development.
> I agree, it had a more community feel about it. Someone commented at the
> time that the Mowbot, which at the time looked like being a standard unit
> based on our discussions, might turn out to be a comprimise by being
> committee driven.

I understand the issues with committees. However the 'eveyone on their
own' concept has been much more idea poor. Sometimes I feel you are the
only real contributor.

I just think if we can define something, anything that folks can get
their hands dirty with, it'll jumpstart the discussion.

But it has a have a resonable curve to it and right now the information
just doesn't seem to be available to do something reasonable...

> >
> >Secondly what about actual grass cutting? My last foray into this field
> >(pun intended) was last spring. Someone on the list has experimented
> >successfully with attaching Exacto Knife Blades to the end of a motor
> >driven platform. I repeated the experiment, using a round plastic J-box
> >cover with the attached blades. I found that if driven by a electric drill
> >it was fairly effective. The only other addition it needed was some type
> >of protection above the disk because the cut grass kept wrapping around
> >the spindle. I envision something like an oil funnel. Cheesy ASCII graphics:
> >Of course finding a powerful enough motor is another problem. I want to
> >thank Paul for posting that motor driver board a while back. BTW is that
> >a kit or is it fully assembled?
> >
> >Can anyone talk about motors, drivers, gears, pulleys, bearings, and other
> >mechanically oriented stuff required to get a platform going? The drive
> >electronics and sensors are rather ineffectual without a platform to
> >carry it on.
> >
> The biggest problem with motors is that the most suitable one for your
> project is always very expensive. Mostly I've restricted my projects to
> what is cheaply available. This is not always a good idea, it's pointless
> designing a great robot if it has trouble pushing itself along.

OK help me here. Let's presume the playfield is low voltage DC motors
(12-24 volts). I know of a two types:

1) Typical small hobby motors.
2) Large motors like winch drivers and car starters.

Neither seem to fit the bill. What other types are there? Where can you
purchase them? How do you interface them to your platform (i.e. mounting,
gears, wheels, etc.)

BTW what is expensive. All the toy motors are cheap. The others?

What about salvage. Do motors from equipment like cordless screwdrivers
or cordless drills fit the bill? They are commodity enough to purchase
and strip.

I'm unfortunately a computer and electronics guy. I'm willing to sweat on
the platform/hardware but I don't know enough to ask intelligent questions.

Every design I see has some set of motors that doesn't seem to be
available. With a little more information I figure I can adapt whatever's
available to my needs.

> For drive motors I chose a wormdrive package with a speed close to what I
> wanted. By choosing a speed close to my optimum I eliminate the need to PWM
> the motor, which simplifies the software. I use differential steering and
> relays to reverse the voltage for driving in reverse, since the robot is
> usually going forward, even when turning, the relays will get used very
> little and won't put much burden on the batteries. The wormdrive means I
> don't have trouble with downhill runs as the motor has it's own top speed
> and if power is removed the worms act as brakes.

OK. Define wormdrive please. What's the RPM, torque, amps, voltage, price,
and distributor of suck motors? What's the drivetrain.

Am I asking stupid questions because this all on the website? I haven't
looked in over a year...

> One problem with direct drive motors to wheels is that going down an
> incline, you can exceed your safe speed (for cutting and stopping). If you
> were normally PWMing the motors to control speed, you now need to PWM the
> dynamic braking input or imitate dynamic braking if your H-bridge doesn't
> have one built in.

I think the bridge that Paul suggested has what's called regenerative
braking. For less than $30 US I think that's a component I'm willing to
buy instead of build.

> On your cutter motor problem. I will be using 2 cutters so I can monitor
> average voltage to determine cut and uncut areas roughly. I have chosen a
> Mabuchi motor that is readily available at hobby stores RS-380PH. The motor
> is small, can run up to 14000rpm and has fair torque (I used the motor in a
> lead screw portion of a walking robot some years ago). If you need more
> torque, you could try the motors used for electric model aircraft.

Aha! That helps. So I need to visit my local hobby shop. WIll do.

And from your comment it's clear you know a bunch about designing
autonomous equipment. Consider me your apprentice....

> >Sorry to ramble. If you're still here you're as interested as I am. What I
> >thought about while mowing over the weekend was how to define a reference
> >point to mowbot development. I finally hit upon the idea of the reference
> >grassbox. The grassbox has these parameters:
> >
> >1) 20 foot square
> I take that to be 20x20 not 4x5.

Absolutely. Sorry. 20x20 is correct.

> assuming a cut width of 12 inches and a pass efficiency of 50% (each pass
> of the robot overlaps the previous cut by up to 50%), the robot needs to
> cover about 600ft2, let's be safe and say 800ft2. To effectively cut all
> this area, it would be neccessary to maintain some sort of map of cut and
> uncut region, this could be implemented as an X-Y grid using odometry from
> the wheels to count distance and compute vector. Odometry is inherently
> error prone, especially on uneven surfaces like your average yard, but your
> 6 inch boundary walls could do some error correcting.

Phweee! Flag on the play! I'm currently uninterested in the intelligence
required to cut the area. It complicates manyfold what needs to be done
first: namely defining a platform of motors, batteries, and cutters than
can do the job autonomously.

Just for the sake of argument let's pretend the platform will run via
remote control. No intelligence necessary for now.

How can a system of motors, wheels, metal/wood, be put together that can
cover the area?

> >2) Bounded by 6 in high boundaries
> bump sensors would satisfy that requirement, and that is the first sensor
> to be implemented on my mowbot.

I put than in precisely because of your bump sensor post. The other reason
is that it takes the pressure off of thinking about how to do edge
detection and other sophistocated sensors.

I'm trying to strip away the smarts and figure out how the muscle works.
How to design a platform that will stupidly and aimless roam the yard
actually cutting something. Adding smarts can be done after it does the

The requirement defines an easily detectable, well defined boundary.

> >3) Sloped at a 40 degree angle.
> If the robot weighs 20kg, then to the motors it may weigh 30kg going up
> such a slope. I don't have the data, but it would seem to have to overcome
> an extra 0.5g to climb the slope and since mass and weight are the same on
> Earth, it would appear to weigh 1.5 times the horizontal weight. My robot
> just climbs a 50deg slope, but it's real slow. Assuming that half the time
> it is going up the slope you will increase current drain by about 3-4. My
> robot draws about 1.5amps on level grass and 4 amps going up about 30deg,
> on concrete the current is much lower.

I put this requirement in for three reasons, two of which you've named:

1) Power consumption
2) Speed.
3) Ability to navigate steep territory without tipping over.

> >4) Contains grass 1 ft high
> This is a hard one. I think you would need to hand cut the area first and
> then use the Mowbot subsequently. 1ft high grass will be a serious
> impediment to any robot, the grass will form a barrier that must first be
> knocked down, then the grass under the wheels can cause slippage which will
> affect any odometric data. As the grass is getting cut it will become a
> thick mat putting great strain on your cutter motors.

Good. This is helpful. So what would be a reasonable maximum height?

If everything works out it won't be a problem. With automation comes
repitition. Once the grass is cut to two inches, a mowbot should have
have no problem maintaining.

> >Given the grassbox here are some performance parameters for a mowbot:
> >
> >1) MB can cut the given area to a 2 inches.
> Can't see too much trouble with that.
> >2) MB must not tip over.
> Keeping a low profile would eliminate that problem and will allow the robot
> to go under low obstacles like branches on a bush.

So a wide footprint and low body. I qas figuring something like a 18 inch
diameter body standing about 10-12 inches high. I'm eyeing the top of
a circular wood table that my wife's gotten for $7 are a prototype platform.

> >3) MB must cut all grass to within 4 inches of all boundaries
> Mine will cut to about 2 inches for a fixed boundary and 0 for concrete
> boundaries.

> >4) MB must perform given task on a single charge.
> That shouldn't be too hard, but we need to know vehicle weight, drive motor
> and cutter motor current draw, wheel size, wheel speed.

I have none of these paramemters, just guesses. Power I'm willing to
sacrifice just to learn how to put something together.

> >5) MB must be buildable using hobby available components.
> That's always been my goal, although there will always be special
> components, like axles or cutter hubs that are best constructed on a lathe.
> >6) MB total cost must not exceed $X (not sure what X is. Subject to
> discussion.
> >Probably would be nice if it were in the ballpark of a good power mower, say
> >$300 to $400 US.)
> HAHAHAHA! Well, we'll see ;-)

Well let's talk about it. Remember I'm talking about just the powerplant.
Can you please help define the cost of motors, batteries, wheels, casters,
and frame?

No electronics are involved. So where exactly is the big cost?

> >Note that the performance parameters only define the powerplant (body, drive
> >and cut motors, batteries) and not the electronics, which can be added later.
> >If anyone can tell me how I can buy/salvage electric motors and connect
> >them to lawm mower wheels, I'd be eternally grateful.
> I had to have axles lathed up. I also had some castings made to take sealed
> barings to take the weight off the motors. I can't see how you'll avoid
> this unless the motors have substantial bearing built in, and I don't mean
> bronze bearings. The casting could be eliminated, I found a flange bearing
> that would do the trick.

That whole paragraph (except for maybe sealed bearings) was greek to me.

> >
> >BTW the last post pointed out using a 500 oz-in ($250! Ouch!) servo. It seems
> >to be a round hole (electric lawnmower designed for human operation) square
> >peg (making it autonomous) problem. Can't steering simply be done by having
> >centered, bi-directional drive motors? That's why I think building a platform
> >from scratch is such an excellent idea.
> Yes it can, in fact centred wheels aren't neccessary. I have my drive
> wheels at the back and a single caster at the front. using the odometry and
> the robot dimensions I can roughly compute the position of the nose and
> sides, it avoids balance problems as I can keep the batteries very low on
> the COG. Ground clearance is 6 inches at the front and 2 inches at the rear.

That's interesting. So you think that hacking on an electric mower is
a possibility, especially for the mechinically challenged, such as myself?

Thanks for such a thoughtful response. I'm just trying to figure out how
to get into the game. That means focusing on muscle and bone, not sensors
or brains.

I'm not sure who's listening to our mailing list, but I'd sure like to get
a sense of how many folks are interested in designing and buildling
somthing simple to start experimenting...