More Mule then Rover?

[Tommy_T]

Now, you could build a G-code interpreter on top of such an approach in order to give motion commands, but underneath it, you'll need a localization routine (especially for something like RoboMagellan, where the distance and terrain covered are sure to impede any dead reckoning approach).

lnxfergy
After reading the rules for the RoboMagellan contest, I can see the above working.

you'll need a localization routine

As it turns out CNC machines have to also, they have sub routines to check the tool with
edge finders, lasers and optical systems to account for things like tool wear.

In my case maybe I could use a G45 [X value] [Feed rate value] this would buffer the current bearing,
then zero the gyro, Acc to Feed rate value using the gyro for directional feedback until math.Abs(Q) < 1 (@3sec),
then Q axis would be used for directional feedback until end of line.

A G46 [X value] [Feed rate value] This would buffer the current bearing, Acc to Feed rate value using
the digital tilt comped compass (Phidget 3/3/3) for directional feedback

The next thing I'll look at is ultrasonics, I see they have some that go out to 200"+
I still have room for 4 analog inputs.

Any feedback on ultrasonic transducers would be much welcomed.

My net-book main controller draws the line at video while checking the feedback devices at 50Hz.

Tommy

[Adrenalynn]

That's hardly localization...

[Tommy_T]

Adrenalynn
Quick question, on an uneven surface would 3 wheel be more stable then 4?
with 4 wheels I can see one rear wheel being in the air at some time which could
cause it to pull that way, where one rear wheel would always be on the ground
(220lb platform) and if it hit a rock I could see it pulling more evenly on both drive
wheels. but I could be wrong.

That's hardly localization...

I should also mention I'm not a machinist.

My day job is systems integration which means I never get to work on any one technology
long enough to gain the experience my age would suggest. With that said, I can see a connection
between a CNC machine counting turns of a ball screw to get to a location and dead reckoning
and then use some type of device or method to re index the physical and software values. In
todays CNC industries they may be looking at +/- .0001" and they can do with high count encoders
and ground ball screws, I only need +/- 2", so I wounder if I could use a similar approach
for my project.

example:
Use high count encoders for dead reckoning then from time to time call a sub routine that would
take the mule to one of four fixed markers(one each corner, or center of lawn?) that it could then
use to correct for any missed counts.

These markers could be rectangular with each side a different length. I did some tests with a
rotary laser($7+shipping e-bay) and a web cam (second picture below). I can also see this
type device giving me feedback on grass height and even whats been cut for minimum overlap.

Picture 3 shows why I like the G-Code approach.
Each of the zones could be a sub routine defined by it's start location, Length and width.
if for any reason it aborts the program, it can be restarted at any Zone, or if my mom wants
to make her deck bigger only them zones would need to be reprogrammed.

A program may look like below
N05 T1; {this sets the offset for a 20" dia. mower}
N010 G90; {sets Abs mode}
N015 G58; {this runs sub routine to align itself to Marker #1 and zeros X and Q axis} this is the main Datum point
N020 G35 [X value] [Time value]; { this turns on the routine to re-index on Marker #5, when X
moves the amount of X value or Time has past the Time value(which ever comes first)}
N025 G01 [X value] [Y value] [Feed rate]; {this would move the mule to start of Zone 1}
N030 M4; {this turns on Mower, and Turns are CW, M5 the turns are CCW}
N035 G41 [X start] [Y start] [X length] [Y length] [feed rate]; {starts sub routine to pocket mow Zone 1}
N040 M3; {turns off mower}



If these ideas have been kicked around a lot making these posts (kicking a dead horse) please let me know.
this is all very new to me.
Tommy

[Tommy_T]

After much internal debate it seems testing would be the best way to
find out if 3 or 4 wheels would be best for uneven surfaces(like a lawn).

the lowest cost would be 3 wheels, thats where I started.

While my first tests on blacktop worked out sweet, the lawn is the real test.

Picture below is the rear caster area of the Mule.

Tommy

[jdolecki]

Hey Tommy

Have you seen this pump?

HYD pump from American Science and Surplus.

For people with bigger bots this seems to be a good deal for 49.00.

http://www.sciplus.com/categorySpecial.cfm

[Tommy_T]

HYD pump from American Science and Surplus.

jdolecki
You may be right, Hydraulics would be a good option for raising and lowering the rear stabilizing wheel,
With the electric drive I'm using which can pick up 300+lb it takes 2sec. per inch at full speed.

As it turns out I'm now looking at adding one more linear movement, Lifting and repositioning Q axis,
I should say I hope to do it with only one linear movement(Thinking V-shaped fork design could lift and turn).


Also I'm now thinking the rear wheel should be a drive wheel, I thought the rear wheel as designed,
would also give me a way of pushing the lost motion in my main drive gears one direction so it could
be comped. but I have to turn to take advantage of it, if the rear wheel was a Drive wheel I could
then push the lost motion while going straight(using a Current feedback loop, I think).

I looked at the American Science pump, but the only stats they list is the line limit Psi @3
I find it hard to believe that @7Amps could only supply 3Psi (I'l look for more PDFs on the pump).

I was just working at a company that made Hydr pumps(SPX), their dumpster was full of different types of
pump parts.

Tommy

[Tommy_T]

Busy week for my PC for work so had little time to work on the software for my Mule
But I did get the front wheels change out.

new wheel 9.5" wide, 17.5" tall
old wheel 3" wide, 14" tall

found these at a place called Tracker Supply at $60us each with the rim

While I only had a few hours to test them out, they are sweet very stable, but they do draw more
current. with the test mower deck installed and a 15deg slope I'm drawing 30amp with the old wheels
it was about 20Amp.

Picture below old wheel next to new wheels installed.

Tommy

[Tommy_T]

Some changes, some playing, some progress.

The broom idea for moving leaves around didn't fly. rotating sweeper seems like the way to go.

Tommy

[jdolecki]

Looks good.

I did see in the Robo Magellan rules there is a weight limit.

So I dont think we willl qualify unless they make a heavy weight / unlimited class?

[Tommy_T]

I did see in the Robo Magellan rules there is a weight limit.

jdolecki
With a 50lb weight limit you'd would be hard pressed to use wheelchair motors that
weigh 15lb each, I changed out my motors for larger four polled ones that weigh 25lb
each, so I couldn't even enter my motors taped to a laptop. I'd think with the 100lb
ballast(2 sand bags) I added my Mule is at 295lb without any attachments.

Looks good.

Thanks for being kind, but looks never was part of my project. but my Mule has been
sitting out exposed to the environment for the last two weeks without any water
damage(Sweet!), we even had a large storm come through last week with 60mph(wind gusts)
and lots of rain(but no snow, which I was hoping for so I can test out my new snow plow).



Do you have any pictures of your current platform?

Tommy