Printable View
Good idea Tommy, I've just ordered a sonar device and IR sensor to play with. Not sure how I will use them exactly yet.Quote:
Originally Posted by Tommy_T
I will probably use pic microcontrollers and the Atmega328, time will tell, lots to learn since I'm new to alot of this. I do like the night time idea.
Hi Tommy_T,
I have a quick question for you:
Correct me if I am wrong, but I believe you are controlling the Mule through a wireless home network. Do you find when you are operating the Mule in a really bad snow storm that your connection to the robot suffers significantly? Also, what kind of usable range are you getting from your connection (both in clear weather and in a blizzard)?
Thanks!
parallaxQuote:
Do you find when you are operating the Mule in a really bad snow storm that your connection to the robot suffers significantly?
I have the Mule's wireless router(NetGear) mounted outside, and yes I have had connection
issues during heavy snow falls, which I figured was caused by the router being covered with
snow(router sitting on roof of house), after I clear the router of snow I'm back to plowing.
Can't give a good answer to that question, because all my wireless range tests have beenQuote:
what kind of usable range are you getting from your connection (both in clear weather and in a blizzard)?
done without heavy snow fall.
As is with the router outside(on the roof) the Mule's PC to router range is 450ft, Mule's network
cameras to route range is 300ft.
Tommy
Thanks for your quick reply! :)
BTW, I am a big fan of your project. It's executed cleanly, has a well-defined purpose, and is simply awesome. Great work!
The reason why I'm asking about these ranges:
I have been in the early stages of designing a general purpose farm-bot for a close friend of mine's family. More like a feasibility test, actually. I'm seeing if some of the things they would like it to do are reasonable at their location. They would love to have a similar plowing/mowing functionality as your Mule if possible.
However I am hitting a range roadblock. The family's driveway is close to 450 yds long (more like a road than a driveway) and other areas of the farm are even further away. I have toyed with setting up repeaters down the length of it, but it seems like a logistical nightmare to try and setup/maintain. After hearing your experience with usable ranges via WiFi, I think I may have to explore alternate methods.
Just a thought~ signals in the 2.4GHz range are severely degraded when traveling through water. Have you ever noticed any performance hit on a rainy day?
I agreeQuote:
I think I may have to explore alternate methods.
it seems the best option are not always the easiest. Wifi while being very easy
to incorporate into most any system, has a number of shortcomings in outdoors
applications.
I gave the Mule the processing power to do it's tasks without the need for me to
look over it's shoulders(network cameras), but because of limitations of the
available transducers or my shortsighted use of them all tests this winter have
been under Wifi remote control.
This week I'm pulling the Wifi cameras and start using the processing power I gave the
Mule to accomplish it's tasks(easier said then done).
Tommy
Is the signal inhibition being caused by direct contact of the snow? If so, encapsulating the router in a transmission transparent material may help? Something simple like a suitably sized plastic cake storage piece with cover comes to mind - some are made so that the cake can sit on the cover for serving. The cover can be secured to a separate platform for easy mounting/dismounting to the roof.
Got the other two test mower mounted, now to look at spring loading.
Tommy
So I guess my post didn't even warrant the courtesy of a reply? I thought it might be helpful and certainly didn't intend to waste our time.
Sorry Stobs! No one means to offend you. I guess your post just slipped through the cracks.
To answer your question:
I'm reasonably sure that the cause of the signal inhibition that Tommy_T has commented on is caused by the encapsulation of the router by the snow and not direct contact. 2.4GHz signals degrade very fast when travelling through water and although snow is frozen, its essentially still just water molecules in a solid state. In this case, using a container like you described wouldn't help much. Good thinking, though!
Thank you very much for the reply parallax - although I'm confused as to why you replied instead of Tommy_T! ^.o
I sent my inquiry as an IM because I didn't want to post anything that could've been construed as inflammatory, but since I sent it to the wrong party I owe Tommy_T an apology for the faux pas and gratitude for passing the message on to you.
I've been following this thread pretty closely and have been inspired by its successes - I'm keeping a similar project in mind for a year or two down the road, as I become more familiar (and hopefully competent!) with robotics.
And thanks for the kind words, even though the idea wasn't really very helpful at all; being given the benefit of the doubt is appreciated! :}
PS: Well, let's make that two of a kind then. O.o I thought I'd sent the above post (#88) as an IM to Tommy_T, so explaining that may clear up some confusion I may have injected here. I hadn't seen that post up there until after I'd sent Tommy_T a "follow-up" IM and then refreshed the page. My additional apologies to you parallax and to the community at large.
Stobs, I'm thinking Wifi is not the answer for this project. yes it was easy to implement
giving me more time to work on weights,power and control issues the Mule's first winter.(I learned a lot)
But it is unreliable at distances greater then @300ft or non-line of sight of @200ft.
I gave the Mule the processing power to accomplish it's tasks, IMO it's time to use it.
Time to start thinking outside the box for ways to keep the Mule in it's box(my yard).
Tommy
I really don't know much about communications, but what about working with cell phones? They seem to work in any weather - well, if you believe the commercials anyway :} and some models have USB connections. The family plan that I belong to lets us upgrade our lines at two year intervals, depending on when we started service/last upgraded; might be able to make a contact at one or more of the local phone outlets and see if you can put up a wanted notice for someone's old phone? That wouldn't work that well with my service plan, but if yours is high/unlimited minutes and reasonable data rates it might be worth looking into.
As is I have a USB 2G/3G aircard(for car's PC) with unlimited air time, but this august the muleQuote:
but what about working with cell phones?
is going to my mom, she would say no thanks if it costs $60+ a month on top of the charging costs.
I'm also shy about connecting the Mule to the Net, all tests with Wifi have been with routers not
connected to the net.
I'm testing creative ways to use a webcam so the Mule can follow a programmed path.
Tommy
Exactly why it wouldn't work for me either, at least not with my current plan.Quote:
Originally Posted by Tommy_T
Completed tests of friction drive Reel mowers on the Mule, and while it had it's advantages
it couldn't give repeatable results, too many things can stop the Reel from turning, which
does makes for a very safe(as mowers go) mowing deck, but requires human intervention
to remove object before it can continue. would work for gulf coarse greens, but not large
lawns with trees or bordering corn fields(like my test area). Also with six fixed wheels in
contact with the ground while mowing lots of energy is needed to maintain a straight cut
path.
Next testing electric motored conventional rotary mowers(I got one last fall)
Tommy
I agree with Tommy, the reel mowers are great as long as you clear the yard of limbs and such before you start mowing.
I have been experimenting on a small scale with the use of GPS. I have had some success but so far the accuracy is not there. I may purchase a more accurate module later if budget permits.
Something I have been running around in my mindseye is using object recognition and have the mower place cones or other objects at the end of each run. The object would be tracked keeping the mower inline with the object. When the mower reaches the object, it picks the object up and places it in the path of the second run. This process would require 2 objects for each end of the yard area. I have used Keyence Cameras with object recognition at work and they work really well. Not sure how well it would work with such a busy environment and with changing perspective.
Anyways I am on hold until the flood waters of the mighty Mississippi have subsided. I have not had water problems yet and probably will not but just in case I have most of my hardware stored in safe keeping until then.
Pabble, I also have been testing a GPS(Phidgets), I can't use a digital compass to getQuote:
I have been experimenting on a small scale with the use of GPS.
the Mule's bearing because of the EMF of the electric motors, both drive and 38" electric
mower deck, I hope to get that information from the GPS.
Not a bad idea, but would try this first(seems simpler) http://www.petsbycsn.com/Loc8tor-LOC...R1001.html?cv=Quote:
my mindseye is using object recognition and have the mower place cones or other objects at the end of each run.
I duct taped a 38" electric mowing deck together.
Tommy
Tommy, take a look at Roborealm. It has the ability to process webcam video and use it to navigate a robot. There is a module available that will record the webcam video of a driven path and use it to drive the robot. The navigation module can also plot its own obstacle avoidance path.Quote:
Originally Posted by Tommy_T
BTW, I am also building a large rover based on a Jet 3 Ultra wheelchair and want to use it for lawn mowing. So I have been following your project with great interest. The mower I'm looking at is the Fiskars Momentum, which uses a flywheel to turn the blades, and is a push not pull reel mower.
Pabble, A good way to test a GPS device would be to record the GPS data(say every second)Quote:
I have been experimenting on a small scale with the use of GPS.
then overlay that data on say a Google Earth image.
In the example below the light colored line is the GPS data, I used a joystick to run the Mule
around the mowing area recording the Lat,Lon data to a file. As shown below trees do skew
the GPS values, while open spaces does a very good.
coppertubing, I have yet to check out any of the ROS packages, all examples I'v seen are smallQuote:
take a look at Roborealm.
indoors robots, hard to see a connection with the Mule.
If your not going to use the P&G motor drive and would like to unload it, let me know.Quote:
BTW, I am also building a large rover based on a Jet 3 Ultra wheelchair
Tommy
Very cool Tommy, I will have to experiment with that.Quote:
Originally Posted by Tommy_T
Were you driving the Mule by camera while gathering data? If so you seem to be getting a very good range and I would be very grateful to know what you are using for a transmitter/receiver. The best I can get by camera around buildings and such is roughly a couple of acres. I get a very good range in open field with no obstacles.
Mostly I enjoy mowing the yard on my riding mower but when it gets windy and dry with all the dust and pollen then my sinuses and eyes suffer. Thinking of modifying one of my motorcycle helmets to wear while mowing hehe:happy:
My Video receiver is hooked to a 32" LCD HDTV and I tend to get motion sickness if I drive by camera very long and mowing remotely gets very boring hence the search for autonomous operation.
Some creative use of Duct Tape, I like it:happy: Looks like two mowers side by side. Are you drawing power from the Mule? How long can you mow per charge?Quote:
Originally Posted by Tommy_T
Pabble, No I was walking along with the Mule while recording the data. I pulled theQuote:
Were you driving the Mule by camera while gathering data?
cameras off the Mule as soon as I figured the chances of snow was over. IMO the
cameras was slowing down implementation of autonomous operation(was a crutch IMO
because the cameras would never give me the range needed).without them I'm forced
to work on the sensors.
Last winter I could spend over three hours running the Mule by camera, I installed wide angleQuote:
Video receiver is hooked to a 32" LCD HDTV and I tend to get motion sickness if I drive by camera very long
lens on all the cameras, that may help.
Thats it, two 20" mowers pivoted 25deg so they overlap with no gap.Quote:
I like it Looks like two mowers side by side
Yes, I removed the batteries that came with the mowers(20Ah).Quote:
Are you drawing power from the Mule?
The Mule can mow for 1.5 hours with the current batteries, I got room for 4 more carQuote:
How long can you mow per charge?
on the Mule when needed(right now I have steel weights in place of those missing batteries).
seems the Mule needs to weigh 500lb for both plowing snow and mowing.
Tommy
Roborealm would be suitable for any size rover that uses a PC for its brains. There are modules for Sabertooth motor controllers (I'm using a 2x25), Phidgets sensors and servo controllers, GPS, and lots more. It makes it very easy to coordinate servos, sensors and motors to provide autonomous operation to your robot. I don't think GPS by itself will allow your mule to navigate accurately enough. You're going to need some kind of vision processing, and that's what Roborealm can do.Quote:
Originally Posted by Tommy_T
No, I won't be using the P&G so make me an offer.
coppertubing, You may be right, but I do have try one other idea thats floating in my head(beacons).Quote:
You're going to need some kind of vision processing
The idea...
The absolute position senor is the beacons, but it limited to only four of them, and it takes up to five
seconds to get a good position fix. to over come the four beacon limit the GPS is used, because it can
track the Mule to within 20ft,that position is used to determined which four to use(my house's four
beacons or my mom's four beacons). The Yaw sensor is used to overcome the five second limitation
of the beacons, the Yaw sensor can track the Mule to within ~.1 inch and .005 deg. at mowing speeds
and uses this information in between beacon updates to help keep the Mule on coarse.
I'm thinking because the Mule has to operate in rain, snow, fog, day, night. implementing a vision
navigation system would be tough.
does it have the connectors in the picture below? and work?. is so, I'd try $50 + shippingQuote:
I won't be using the P&G so make me an offer.
Tommy
Good point about the vision processing in all weather conditions. Although I haven't tried it, it seems like Roborealm can do all manner of video processing, which may allow it to work in those conditions. I know it can track a laser dot, and it has a module for the Kinect which can use infrared cameras for obstacle avoidance. Once I get to that stage, I'll see what it can handle.
As for the controller, your plug looks like it is for the Jet 3 - this one. Mine is the Jet 3 Ultra. Don't know why they're different, but this is the controller for it. Since they go for $600 new, I guess I'll put it up on ebay.
coppertubing, if you get no bidders keep me in mind, I do have a sabertooth 2X25(Never used) That I would trade.Quote:
I guess I'll put it up on ebay.
Tommy
Mule's first try at mowing in a confined area in CNC mode
http://www.youtube.com/watch?v=-9qfBXX8_Rg
Tommy
While I expect years of tinkering ,fun and use of the Mule, this project is complete.Quote:
Tasks it 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)
time for the next project.
Tommy
Props to you for this project, Tommy (I had followed you on SoR, but was delighted to stumble upon the lengthy thread here :) )
I got ahead of myself.Quote:
time for the next project.
While yes the mower deck does cut grass, it sucks!. very under powered, very noisy and too big.
my first try was using three reel mowers in their stock friction drive configuration, I'd pick them
up when making turns,then lower them down to mow. this setup cut very nice, was small and
quite, but small sticks, corn cobs or barbie dolls would stop the mower causing the mower's
wheels to slip.
Next I tried stock battery powered electric rotary style mowers(Earthwise cordless 24vdc)
but as stated above, they suck for my application.
So my next option seems to be putting motors and casters on the reel style mowers.
Tommy
Ever try adding a small grating in front of the reel mowers to catch debris and push it to the side (think cattle catcher on a steam locomotive)? Or maybe an auger, like on the front of a combine, to direct debris to the sides? Or invert the auger to direct debris to a small conveyor belt in the center to collect debris into a container (pine cones friggin' hurt when they make it out from under a rotary lawnmower)? Or a 'jamming' gripper on a small-ish arm to grab debris sighted by a camera observing its path (collect or throw away)?
Can you tell I've wanted to build a robotic lawnmower? I had once considered using a reciprocating-type clipper akin to a hedge trimmer, but that seems to have the worst parts of both a reel and rotary mower (still needs some sort of protection like a reel mower and a power source like a rotary mower).
tician, good idea!Quote:
Ever try adding a small grating in front of the reel mowers to catch debris and push it to the side (think cattle catcher on a steam locomotive)?
I have not tried so far, one reason a rotary style mower cut so good is it sucks the grass up as it's cutting,
a rake in front of a reel mower may help stand the grass up for a better cut, and help keep the barbie dolls
out of the blades.
Tommy
Would seem I have a design problem.
in last two years I'v broken each of the drive axles once, the axle is 1.000" dia. steel turned to .800" dia. for the
bearing, each failure occurred at this turned surface. the first was in the middle between the bearings, yesterday
it was the other side at the output edge of the bearings.
I'll change the bearings to 1.000" ID dia. and use retaining rings, in place of the turned .800" shoulder.
Tommy
You've already got a fix lined up, but what was the corner radius at the shoulder? Sharp internal corners are great stress concentrators.
The first failure sounds interesting though, did you figure out a cause?
Any sort of surface defect on a shaft will become a stress concentration point, and not just the shoulder - although the shoulder radius is an extremely important parameter during design. Get any sort of wear in one location of the shaft, and odds are that will be where it fails.
What really stands out to me as a likely cause of failure is directly supporting the wheel with the output shaft. Shafts are intended to withstand the large shear stress of torsion, and having the wheel cantilevered so far out on the shaft will cause significant bending (tension/compression) stress that they are not so capable of withstanding.
Another question to ask is: what type of steel were the shafts? Because not all steel is created equal.
Th232, it seems when I made the bearing blocks I didn't make sure the bearings were fully seatedQuote:
The first failure sounds interesting though, did you figure out a cause?
(press fit), one bearing was cocked a little, over time it caused the axle fail.
.020" is the corner radius the failure happened @.090" from the corner radius.Quote:
but what was the corner radius at the shoulder?
tician, the shafts used were 1045 steel with a hardness of brinell 150(Rockwell 80).Quote:
what type of steel were the shafts? Because not all steel is created equal.
when they were assembled there was a slip fit between the bearings and the shaft, now there seems to be .012" over on the outside bearing.Quote:
Get any sort of wear in one location of the shaft, and odds are that will be where it fails.
Tommy
The reason I don't suspect the shaft steel or wheel placement.Quote:
What really stands out to me as a likely cause of failure is directly supporting the wheel with the output shaft.
The wheel hub shoulder is .670" dia. , the bearing shoulder is .800" dia.
none of the failures occurred at the smallest dia.(wheel hub).
I believe all the failures are a result of the bearings, or bearing fit/installation.
on a side note:
if you gotta cut grass with batteries the reel style seems to work best.
Tommy
tician, I'm being to see it(I get slower as I get older).Quote:
What really stands out to me as a likely cause of failure is directly supporting the wheel with the output shaft.
each of the failures showed excess wear on the shaft at outside bearing, with no measurable wear on the inside
bearing. the bearings are rated at 2900lb, even with 140lb of sand bags(when wet heavy snow) the load on the
wheel is under 350lb, but because I'm powering the wheel(motor sprocket) on the same side as the load carrying
wheel it is causing excess wear at that bearing(outside bearing).
I increased the dia. at that bearing from .800" to 1.00" and I'm beginning to see all I did was postpone the next failure.
A better fix would be to make room to move the motor sprocket to the other side of the spindle(inside bearing side)
in hopes of evening out the wear between inside and outside bearing surfaces.
Tommy
Honestly, this thread has been stuck in my head since I first read about the shaft failure. Shortening the shaft so that the wheels are closer to the frame should help increase the life a bit as it should decrease the bending stress a bit (the bending forces are what causes the bearing to eat into the shaft). Ideally there would be a space frame of some sort to connect the wheel to the bot's frame (to support the weight and forward forces) instead of using the shaft to do it all. This would permit a smaller diameter shaft since it would no longer be supporting the weight of the bot, just the torque between the drive sprocket and the wheels (with a not so small radial force caused by the drive sprocket). If I have time this weekend, I may break out the mechanical engineering textbooks and notebooks and do the math (mostly been EE the last while, so the ME has suffered).
If I am thinking correctly, the axle can be approximated as a cantilever beam. The wheel end of the shaft has an upward reaction force equal to the weight of the bot applied at the bearings, the forward reaction force caused by the wheel (due to the torque applying a force at the ground via the wheel), and the reaction to the applied torque. The wheel end is unconstrained (free to rotate and translate in all dimensions). There will be a very small bending moment at the end of the shaft when the weight of the bot causes the wheel to canter inward (caused by the reaction force of the weight not being applied directly to the end of the shaft, but slightly further out), but given the width and conformal nature of the pneumatic tires you are using, I think it should be safe to assume this bending moment is negligible.
The bearing end of the shaft is fixed in translation three dimensions and fixed in rotation in two dimensions. It will have the downward force of the bot at the bearing and a backward reaction force to counter the forward force from the wheel end. Because the bearing end is fixed in 5 degrees of freedom, there will be reaction moments to counter the bending moments caused by the two forces at the wheel end (the upward reaction force to the weight and the forward force to drive the bot). The longer the shaft and/or the larger the reaction forces at the wheel, then the larger these reaction moments at the wheel (higher stress) and the larger the deflection of the wheel end (the edge of the bearing will bite into the shaft as it deflects). This is of course ignoring the torque and reaction force from the drive sprocket, but its location will affect where these forces are found.
tician, because I'm using a 42" wide snow plow, with two wheeled prepositional steering, shortening the wheelbase would haveQuote:
Shortening the shaft so that the wheels are closer to the frame should help increase the life a bit
an adverse effect on controlling the plow's path.
the surface for the bearings is 2.75" wide, the ball bearing surface used is 0.5"(1.0" total for both bearings), if I used roller bearings
in place of the ball bearings, it would increase the surface used and in turn reduce the wear at them locations.
lucky for me when the shaft failed it was near my place(50ft), if it was further away recovering the device would have been very
difficult because of the weight of the device. I need to come up with a recovery option(other then calling to tow truck).
maybe something like picture below.
Tommy