First personal project.

[Tommy]

Hello Forum!

I'm new to both this forum and to personal robotics. for the last 25 years I've worked as a
system integrator for the manufacturing industries and have worked on projects with industrial
robots, but this is a first try at this type of personal project.

I should also let you know that my grammar, spelling skills and artistic talents are very limited.
I also love it when I'm told it can't be done.

My Project: General Lawn care robot.

Project budget $2500.00

Because I still have to work I'm giving myself 1 year to complete the project.

Tasks is most perform
1) carry payload of 150lb (at least 2 bags of potting soil)
2) mow grass (both manually(remote control) and autonomously(with supervision)) 
3) clear snow for drive and walkways manually(remote control)
4) simple to control (even my 75yr old mother can use in her garden)


The main controller is going to be a net-book PC(or my dual core laptop in need be)
remote control operation will be Bluetooth(wireless keyboard) and WiFi.

Servo control is analog(0V to 5V).
Speed controller RobotQ Ax2550 - (Sabertooth 2x25(for testing) and Peripheral Attachments )
Transducers Phidgets 3/3/3, Phidgets 8/8/8, Phidgets high speed encoder cards(at least 2) ,at least to encoders.     
Analog output Labjack U3-LV

After searching this forum on similar project I'll use a used power wheelchair as a test bed(found on craigslist, picking up Saturday).

Feel free to correct me when you see I'm wrong.

Tommy




 

[Conscripted]

Sounds like a useful contraption when finished. I have one question. Will two drive wheels and two casters give you the traction you are looking for to shovel snow or traverse a garden?

Conscripted

[Tommy]

Will two drive wheels and two casters give you the traction you are looking for to shovel snow or traverse a garden?

Conscripted
In my head I can see me having to make and install chains on the Drive wheels or
maybe Tracks, if I can get the test bed at the price quoted on graigslist and it well work
I should then have the budget to get tracks(if needed or usable (could cause damage to lawn or garden?)
A company call ariens makes a Track kit for snowblowers that could be retrofitted.

It seems there are lots of options when it comes to tires.
http://www.wheeleez.com/

Tommy

[Tommy]

I got the test bed (came as advertised, and more..Sweet!)

While disassembling I got a look at the drive that came with it (Penny&Giles 50/70 Amp VSI speed control).
And I gotta say it is a sweet drive, and if theres a way of using it I could shave $500 off the budget.
The other sweet thing about that drive is it has built in safety features like battery level inductions, differential
analog monitoring ,analog voltage level monitoring and Motor current monitoring.(shorted or stalled motor can't take out a drive..sweet)

using the P&G drive would remove the need for a Microprocessor as a hart beat also saving $100.

But it has built in acceleration/Deceleration loops that would cause me issues while under autonomous control.  

First question: Has anyone ever tried to remove or change the Speed profiles on a Penny&Giles 50/70 Amp VSI speed control ?.

I goggled it and found references to a Speed profile programmer for Dealers, OEMs and Manufactures, but none for end users.


Tommy  

[Soeren]

Hi,

using the P&G drive would remove the need for a Microprocessor as a hart beat also saving $100.

A µcontroller is just a few bucks, but I don't see why one motor would need it while another wouldn't?

But it has built in acceleration/Deceleration loops that would cause me issues while under autonomous control.  

Why would it?
Just use encoders on either the wheels or the motor shafts and count pulses, then you know how far each wheel traveled. With a bit of testing, you will probably find how much you need to modify a given distance at a given speed for complete travel.

I should then have the budget to get tracks(if needed or usable (could cause damage to lawn or garden?)

Why would tracks damage the garden more than wheels (as long as you don't rotate too much in one place)?
Tracks will spread the weight over a larger area, so it will be lighter per square area of contact.

[Tommy]

A µcontroller is just a few bucks, but I don't see why one motor would need it while another wouldn't?

Soeren
If the main controller(PC) shuts down for what ever reason and the DAC voltages go to 0v with the Sabertooth or RobotQ
Drives will go Full reverse and full CCW turn until the batteries drain, drives fail or the test bed ends up in Cleveland.
With the P&G drive if the main controller shuts down and the DAC voltages drop to 0v the drive will shut down and show
a fault. also if I power up the drives with the main controller(PC) turned off the Sabertooth and RobotQ will take off, but
not the P&G drive, it would show a fault.

It reminds me of the mid 80's when working on the positioning feedback devices on the first generation CNC machines
you always had one hand on the E-Stop button when you first turn it on, because if you made a mistake on the feedback
devise the machine would take off as fast it can until it hit and end of it's ball screw, turning on the test bed with the
sabertooth drive installed and no micro as a hart beat, I hold my breath and have a hand on the E-Stop.

With the P&G drive it's kinda like todays CNCs, if anything is wrong it just faults out.

Why would it?
Just use encoders on either the wheels or the motor shafts and count pulses, then you know how far each wheel traveled. With a bit of testing, you will probably find how much you need to modify a given distance at a given speed for complete travel.

Because I'm not doing random cutting paths(I hope) any deviations from the target path will cause me a problem, with the Sabertooth
and RobotQ drives even if one wheel runs over a toy the kids left laying around or even Dog crap(my mother has a dog) the drive
can make the speed and direction correction very fast(full reverse to full forwards in .5 seconds), because of the Acc/Dec programed
into the P&G drive it takes 6 seconds to go from full reverse to full forwards.

.5 sec = @2.5" linear travel
 6 sec = @30" linear travel
mowing deck on test bed 20", I want it to be 48" when project completed.

Just use encoders on either the wheels or the motor shafts and count pulses, then you know how far each wheel traveled.

because of the reason listed above(dog crap, toys,gopher holes) I can't use the drive wheel in the velocity or positioning loops

Why would tracks damage the garden more than wheels (as long as you don't rotate too much in one place)?
Tracks will spread the weight over a larger area, so it will be lighter per square area of contact.

I was thinking about all that friction while turning, but also how much wider and long they are, and can my mom plant her carrots
without running over the ones she already planted?. I have never used tracks before so my information in very limited.
I willing to give them a try.


Tommy

[Soeren]

Hi,

If the main controller(PC) shuts down for what ever reason and the DAC voltages go to 0v with the Sabertooth or RobotQ
Drives will go Full reverse and full CCW turn until the batteries drain, drives fail or the test bed ends up in Cleveland.
With the P&G drive if the main controller shuts down and the DAC voltages drop to 0v the drive will shut down and show
a fault. also if I power up the drives with the main controller(PC) turned off the Sabertooth and RobotQ will take off, but
not the P&G drive, it would show a fault.

OK, plus one to the P&G then.

Because I'm not doing random cutting paths(I hope) any deviations from the target path will cause me a problem, with the Sabertooth
and RobotQ drives even if one wheel runs over a toy the kids left laying around or even Dog crap(my mother has a dog) the drive
can make the speed and direction correction very fast(full reverse to full forwards in .5 seconds), because of the Acc/Dec programed
into the P&G drive it takes 6 seconds to go from full reverse to full forwards.

.5 sec = @2.5" linear travel
 6 sec = @30" linear travel
mowing deck on test bed 20", I want it to be 48" when project completed.

You wouldn't need a full reverse in such cases, just minor adjustments and when mowing the lawn, you won't be going fast anyway, so I'll still recommend encoders (they could be mounted on two springloaded casters for non-slip).

because of the reason listed above(dog crap, toys,gopher holes) I can't use the drive wheel in the velocity or positioning loops

I don't think a turd here and there should upset precision, but that would depend on which sensors you use of course.

I was thinking about all that friction while turning, but also how much wider and long they are, and can my mom plant her carrots
without running over the ones she already planted?. I have never used tracks before so my information in very limited.
I willing to give them a try.

As long as you make all of it yourself, you decide how wide they're gonna be and of which material.
However, I don't think tracks will be a necessity - most people that makes autonomous lawn motors use regular wheels and if the winter means too much slippage, a set of modified (shortened) car chains should help. With cybernetics, neither snow, ice or gigantic turds should be a problem, as it is part of the autonomy to negotiate such obstacles.

If you can post sharp photos of the control electronics and what else materials (schematics, descriptions, links etc.) you may have, we could perhaps find a way to disable the acceleration/deceleration profiles.

Perhaps you should take a look at the CMU cam for navigation purposes.

[Tommy]

You wouldn't need a full reverse in such cases, just minor adjustments and when mowing the lawn

Soeren
While on blacktopped pavement your above statement is correct, both drives can do a very good job because only small
corrections are needed, but not on my lawn with it's tree roots, 25deg slopes and uneven surfaces. the P&G with it's Acc/Dec
can't even make it one pass while the Sabertooth can. as it turns out not even the 25Amp sabertooth is up to the task.
The 150Amp RobotQ(don't have yet) should do the trick.

so I'll still recommend encoders

I do have two encoders for closed loop control 1-positioning, 1-Directional velocity, no feedback on the drive wheels(yet)
the RobotQ does have 2 encoder inputs for wheel encoders.

As long as you make all of it yourself, you decide how wide they're gonna be and of which material.

If I was going to add tracks chances are they would be  http://www.snowblowersdirect.com/Ariens-72600500/p829.html?utm_source=froogle&utm_medium=shop+portals

Tommy

[blackbeard]

nice idea but it seems a bit under engineered. first off from your diagram it shows casters. casters are a death trap on grass and dirt. now think about this. is it worth spending 2500 on a robot that you have to remote control to cut grass when you can get a riding mower for 1000? also why would you have a wheelbarrow sort of thing if you don't have anything to fill it with? i mean if you have to fill it then it's a little superfluous. lastly i think you could do that with half the budget if not much, much less depending on what you've already spent. a wheel chair is a good platform but you should really think about how to utilize as much as possible and you should realize that they are NOT for driving in the dirt. in fact you will need to change the wheels to something bigger if you intend to use it at all.

[Tommy]

nice idea but it seems a bit under engineered.

blackbeard
I'd have a hard time arguing that point, truth be told this is my first project of this type
and with no prior experience I'm learning as I go, for me learning new things comes easiest
when working on an application. I picked the project to help my mother maintain her independence
in small ways around the yard/garden, and for me to have some fun making it work.

first off from your diagram it shows casters. casters are a death trap on grass and dirt.

While the casters are working even on my sad excuse for a lawn they do create uneven forces causing
automatic control issues, the new test bed I picked up has four casters (2 in front, 2 in back) and they
are wider, worst case would be to have to pull the motors from one test bed and mount on the the second
one for 4 wheel drive.

 

is it worth spending 2500 on a robot that you have to remote control to cut grass when you can get a riding mower for 1000?

With what I have now, I could use WiFi to network the testbed's PC with my mom's PC in her living room
and she could mow or clear the snow from her drive and walkways by cameras. I'd pay much more to give
her that kind of independence.

also why would you have a wheelbarrow sort of thing if you don't have anything to fill it with?

The payload dump box is so she can haul potting soil and other things around her yard, she may not be able
to Handel a wheelbarrow by herself anymore but she can use a joystick to move things around her yard and garden.

lastly i think you could do that with half the budget if not much

Right now I have about 1300.00 into the project, my lover could tell you to the penny, but she's not here now.
The big ticket items still needed are the 48" mowing deck, I have a 20" for testing, and the tracking system for
autonomous functions, I'm leaning towards multiple wireless dog fence transmitters, they advertise real time range
information. even if they have values that could change over time due to sun spots or god only knows what, I
could have a master unit in a fixed location to use as an offset for the rest of the units.

 

in fact you will need to change the wheels to something bigger if you intend to use it at all.

The new testbed I picked up has knobby tires, but if need be I would add tracks.

Tommy

  

[Tommy]

I think I have just fallen in love with the new MK5 Tac drive I got.
http://www.invacare.com/doc_files/1114808.pdf

with linear actuators outputs, limit switch inputs, momentary switch inputs for directional control
,E-Stop Inputs the thing even has a tile sensor input. and best of all it has an external programing
unit that can be found on e-bay or craigslist.

This is one sweet drive, if it works (not powered up yet).

Tommy
  

[Tommy]

Had some time to play with this project, and have come to the conclusion that this type project is very challenging.

The most challenging being the mowing. I hoped to have it mow straight lines as programed, that has not
happened as yet.

Will two drive wheels and two casters give you the traction you are looking for to shovel snow or traverse a garden?   

As i'v come to believe is stability is move important then traction(Feedback not on Drive train).

The first picture below is the original wheel setup(with the test mower installed).

The next one is an test to see if a fixed wheel in the back would help keep it stable on uneven surfaces, It
did work good after alignment adjustments. but it toke 4sec to raise and 4 to lower.


The last picture is a test to see if wider tires would help stabilize my Mule. It did work sweet, much faster
then raising and lowering a fixed wheel. While my mom's lawn is not to bad, my lawn which I'm using for
testing is very bad, complete with 30deg sloops,and drainage ditches(I should take the shovel away from my lover),
To my surprise the Mule spines even the wider tires when climbing 20deg on wet grass. So now I'm thinking about
adding a third drive wheel(replace the back casters).


Tommy
   

[z.s.tar.gz]

With this type of project (especially as a first project) you're looking at a lot of work.
More importantly than any aspect of engineering is the idea that it will work. Maybe not today, maybe not tomorrow, maybe not this month. But it will eventually be finished.
Many people fall into a state of "permanent hold" when they come across a large roadblock after spending a lot of time on a project (myself being a prime example).

And after all, the journey is more enjoyable than the destination.

[Soeren]

Hi,

To my surprise the Mule spines even the wider tires when climbing 20deg on wet grass.

With the mass being equal, the wider tires, while having a larger footprint, have less mass/square" of footprint.

There's not much difference between eg. 2 square inch of 10lbs/sqr" and 10 square inch of 2lbs/sqr" and if anything, the wider tires will slip more easy on wet grass if it goes at any speed beyond a slow crawl.
(Knobby tires or the like that digs in is another matter).

Perhaps if you pair up a combined drive wheel and lawn aerator...

[Tommy]

With the mass being equal, the wider tires, while having a larger footprint, have less mass/square" of footprint.

Soeren
While I can see the logic of the above line, there must be some variable overlooked. Because it seems like the new
tires draw @30% more current then the old tires, in my head I see that caused by greater friction between the
ground and tire, which may also be the reason it makes the Mule more sable on uneven surfaces(which is a good thing).

I don't have drive current feedback my only justification for the 30% value is by testing with different
motor drive power input fuses, like install a 10Amp fuse then perform a task that would make the fuse fail then
increase the fuse value until the task is performed.(gone through 3 boxes so far at @3.00 each box, seems I
should install feedback soon).

My next effort is to replace the 2 pole drive motors with 4 pole, in hopes of getting better low speed torque
in hopes of keeping the tire from slipping. it also gives me the opportunity to open up the gearboxes and
remove the lost motion before they are installed.

Perhaps if you pair up a combined drive wheel and lawn aerator...

Soeren, if I installed something like the picture below, my mom would start to cry just looking at it.
and I wouldn't get too close either, these big testbeds can be a handful.

Tommy

  





  

[z.s.tar.gz]

If you want maximum performance you have to cover everything with spikes, not just the tires.

[Soeren]

Hi,

While I can see the logic of the above line, there must be some variable overlooked. Because it seems like the new
tires draw @30% more current then the old tires, in my head I see that caused by greater friction between the
ground and tire, which may also be the reason it makes the Mule more sable on uneven surfaces(which is a good thing).

It was a generalization. You also get better grip in snow and ice if you let out some air on your car tires and the tire pattern can have a lot to say.

I don't have drive current feedback my only justification for the 30% value is by testing with different
motor drive power input fuses, like install a 10Amp fuse then perform a task that would make the fuse fail then
increase the fuse value until the task is performed.(gone through 3 boxes so far at @3.00 each box, seems I
should install feedback soon).

+1 on that.
or at least a variable resettable fuse or current limiter.

My next effort is to replace the 2 pole drive motors with 4 pole, in hopes of getting better low speed torque
in hopes of keeping the tire from slipping. it also gives me the opportunity to open up the gearboxes and
remove the lost motion before they are installed.

"Remove the lost motion"??
If it's lost, how do you remove it?

Why not gear down for less speed.

Soeren, if I installed something like the picture below, my mom would start to cry just looking at it.
and I wouldn't get too close either, these big testbeds can be a handful.

Yeah, I totally get her, I cry too, whenever I remember that there is never gonna be another Mad Max movie - hey, why not kill two birds with one stone...

[Tommy]

It was a generalization. You also get better grip in snow and ice if you let out some air on your car tires and the tire pattern can have a lot to say.

Soeren
Fact is I'm not all too concerned with the tire slipping issue, because my mom's lawn is not as bad as mine, and
I have feedback that keeps the mule on course even if a tire slips.

"Remove the lost motion"??
If it's lost, how do you remove it?

Each of the four motor/Gearbox that I have(each about six years old) have some Backlash in the gears which
equals about a deg of lost motion between the Mule(Feedback) and the ground(Wheels). this is something that I need to
address one way or the other, The Mule's Q axis has .006deg resolution but I can't use it as is. I currently reduce
the gain of the directional loop to keep it from becoming unstable(oscillating).

Why not gear down for less speed.

Very good idea, if I didn't have the larger four polled motors already which I think would be much easier to install
then try to gear them down, but it may turn out I'll have no chose but to gear down the gearbox.

hey, why not kill two birds with one stone...

Funny you should say that, I was just thinking how to best use the linear actuator I have(gear motor/Ball screw type).
Now it's used to raise and low the rear fixed wheel, but with the wider tires I installed that is no longer needed. I could
use it for directional control of a third wheel(replace the rear casters with a single steerable fixed wheel) , or use it as
a kickstand for changing tires etc../get unstuck in the snow device.

Picture below is a example, if something like that was mounted under the Mule(forward of the front drive wheels)
in a way that when operated would lower the arm and lift the front drive wheels off the ground. If it was operated
longer it would push the Mule backwards some amount, at it's max travel the joint in the middle of the lift arm
would pivot letting the drive wheels back on the ground and also letting the arm return to it's default position
without returning the Mule to it's stuck position. the spring would keep the working end of the lifting arm out
of the way when it's in it's default(up) position.

I could see a device like below with two lifting arms one foot apart.


Tommy

[Tommy]

Some dinking around.

Got the bigger four poled motor gearboxes tightened up(removed 90% of lost motion), but not all that
happy with the way I ended up doing it(drilled and pinned the main output shaft). I also got the motors
installed on the mule, which turned out oK but I'm also unhappy with the way it ended up being done,
(one wheel hub would not come off, so I toke it to an auto repair shop, they heated it until red hot...ouch!
but still could not remove it, I bough a 7ton wheel puller($36us at Tracker Supply) and used 4ft lever and got it off).

While I was at the local hardware store the first thing I spotted was them removing lawnmowers and weedwackers
and putting snowblowers and shovels in their place, which got me thinking about snow removal.

Without having any snow removal by plow experience I figure I should make my own snow plow for the Mule.
that way I could make changes cheaper and faster.

kinda like reinventing the wheel I would think, but it should be fun.

picture below is the starting the first ideas I had(all the parts I had laying around).


Tommy
 



 

[Gertlex]

Looks goood.  I applaud your work, and laugh at the naysayers.

[Tommy]

Looks goood

Gertlex, Thanks but I'll be the first to admit my Mule is all ghetto at this stage, and it changes daily.


Tommy

[Soeren]

Hi,

Without having any snow removal by plow experience I figure I should make my own snow plow for the Mule.
that way I could make changes cheaper and faster.

It sure seems quite a bit shorter than I figured it, but that apart, for a snow plow, you need to slant/angle the blade to get the snow out of the way, or it will just pile up a wall in front of it an not be able to go any further.
Alternatively, a wedge shape can be used (and will help it go straight, but it's only really useable if you plan to run only single tracks. For clearing more than the width of the showel, a slanted design works better.

For proof of concept, an all plywood approach is fine, but a bit of metal underneath it will make it last longer.


Here's some inspiration:
http://www.robotshop.com/blog/roboplow-snowplow-robot-311
http://www.treehugger.com/files/2009/01/worlds-first-solar-powered-autonomous-snow-plow.php
http://dvice.com/archives/2007/02/roombalike_snowplow_robot_uses.php
http://hackaday.com/2010/02/01/snow-blower-robot/
http://www.motherearthnews.com/Do-It-Yourself/DIY-Human-Powered-Snowplow.aspx
http://www.motherearthnews.com/Do-It-Yourself/DIY-Human-Powered-Snowplow.aspx

An alternative "showel"
http://conceptpop.com/radio-controlled-robotic-snow-plow

[Tommy]

Soeren, Sweet links. much to learn from them.

1) Slanted blade the way to go(as seen in first link).
2) Picking up snow may not be best(as seen in second link).
3) Using current tech in wheel tread is way to go(as seen in first two links).
4) Gyro drift could be a tool for programing(as seen in second link)
5) Wide angle lens for cameras, while not used in first link can see how it would help.

This week I hope to change out the rear casters for a bigger single casters(for testing), get started on the body cover(umbrella, very ghetto), and Also hope to have the test plow made.

For proof of concept, an all plywood approach is fine, but a bit of metal underneath it will make it last longer.

The design I pulled out of my..., has three 1/2" steel rods bent in the shape of a ski (left,right and center) as a way to keep
the wear surface for being damaged from sharp transitions (like uneven sidewalk tiles) . but unlike most of the links you supplied
I'm thinking of at least giving not lifting the plow a chance(fixed spring loaded), with zero deg. turns I think I may be able
to pull it off.

Tommy

[Tommy]

With 10" single caster and start of umbrella(1/4" plywood frame, waterproof nylon cover).

Tommy

[Tommy]

Bad week for my Mule, good week for my boss.

Got nothing completed this week, and had storm come through that could have dumped
some snow on my lap(funny, I'm now looking forward to snow?). while at the hardware
store looking at better casters(10" ones I installed are crap), I seen some 30Ah batteries
half the size of my current batteries yet weighed 2/3 as much, thinking I could get more
weight over the drive wheels(as much as possible) with four smaller batteries then two
larger ones.

went with 12" caster for rear of Mule, and much better quality(cost three times as much)
after playing with this project these last few months I thinking I'm going to have to adjust
the budget.

I did get the yellow flashing light installed(all snow plows seem to have them), I got an LED
model that I had to tone down(too bright).

PS: Last week Mule sat out exposed to the environment, had high wind(60Mph gusts) and rain storm
with no water damage(Sweet!).

Tommy

[Tommy]

While waiting for some snow to play in I started playing with the fallen leafs, I had
to remind my lover leaf moving was not part of the original project outline and should
not be included in the current budget, not sure she bought it(but she didn't record the costs, so it's all good)

Tommy
 

[Tommy]

After having a chance to at least look at leaf removal I'm thinking a tilted rotating brush setup is the
best option, with casters on each end and motor driven I should still able to do Zero Radius turns.

Back to snow.
While driving to a job last week I spotted an ATV/Snowmobile shop that had a plow on an ATV sitting
out front by the road, I had to stop and look. after talking to the salesperson I was sold on the tapered
blade setup(less tilt needed), and a ploy blade that weighted 20% less then the steel blade, it also looked
easy to Mod. with a saw and welder(steel frame around ploy blade).

I still see the single rear caster as the weakest point(going to have to look at that soon).


Tommy



 

[Tommy]

Tested single rear caster idea, while it worked sweet on hard flat surfaces like blacktop,
did very bad on uneven surfaces like lawns. did some tests with two 10" casters and it
did much better on the lawn, but at about seven miles per hour the casters start to
oscillate, should have something to do with the angle of the caster yet it measures out
good(parallel to the ground).

Why not gear down for less speed.

Soeren's idea seems even better after pulling around a 42" lawn sweeper, while it
had no problems pulling it with slops under 10deg, over that it would slow down, at
35deg it would stall, I can see no reason it needs to move 20Mph, it's current max
speed.

Tommy




 

[Tommy]

Got the smaller motor geared down.

#40 chain with 10 teeth on the motor, 30 teeth on the drive axial.

Tommy
All opinions expressed are mine alone, and subject to change as information is gained.

[Soeren]

Hi,

Good work overall and particular with the lathe work.
I really enjoy following the progress