Mostly saying hi, and showing off some work :-)

[futmacl]

Hi folks,

So, I'm new on the forum, but I lurked around societyofrobots.com for a while - so I mostly wanted to say "hi". A bit about myself: I'm a hobbyist robot builder in California - my most recent project is this:

http://lcamtuf.coredump.cx/tinybot/

I also authored a ridiculously detailed CNC robot building guide, which may be to interest of several people here:

http://lcamtuf.coredump.cx/guerrilla_cnc1.shtml

Cheers,
/mz

[dunk]

hi mz,
great work!

i have a question about your bot:
are you getting a reasonable turning circle with a solid back axle?

my only ackerman steering robot only had a single rear wheel for this very reason.
after that i switched to skid steering methods because it makes path planning far more intuitive.


the CNC guide is of particular interest to me.
i'm in the process of building a small CNC mill from scrap that i hope will have the precision to mill PCBs.


dunk.

[voyager2]

Nice robot,
I'd love to see it for sale!(i want one)

[futmacl]

i have a question about your bot:
are you getting a reasonable turning circle with a solid back axle?

Turn radius is about 15 cm in this configuration, so it depends on your definition of "reasonable" :-) With a differential drive, I would need the robot to be a whole lot shorter to still be able to get at least some movement accuracy (but that sucks); or use a pivot / omni wheel, but that's difficult to instrument for traction sensing & dead reckoning.

[amando96]

Dayum 

How did you make the gears/gear boxes, and wheels?!?!?? perfect.

[futmacl]

How did you make the gears/gear boxes, and wheels?!?!?? perfect.

That's covered in that guide I linked to (part 2 has a whole chapter on gear geometry, resin casting, etc). It proved to be really painful to learn all this, and there's some up front investment on a CNC mill, but in the end it's worth it I think :-) Certainly compared to mail order fabrication...

[Admin]

After skimming through your tutorial, I'm still a bit confused how you machined your robot.

It appears you made plastic molds for a liquid plastic, but its not very clear given only the 'positive molds' in the image . . .

It looks interesting!


ps - I'm still amazed you made gears and got them to work properly . . . did it work on the first try? Sometimes I can't even get screw holes to align on my first prototype

pps - I got your email . . . may take a few days before I really have time to get back to it . . . busy busy busy!

[futmacl]

After skimming through your tutorial, I'm still a bit confused how you machined your robot.
It appears you made plastic molds for a liquid plastic, but its not very clear given only the 'positive molds' in the image . . .

Yeah, I have it illustrated a bit better on Flickr for an older project of mine:

http://www.flickr.com/photos/candy-bottoms/sets/72157612928904793/detail/

It goes like this:

1) CAD, CAM obviously:
http://www.flickr.com/photos/candy-bottoms/3224851187/#in/set-72157612928904793/
http://www.flickr.com/photos/candy-bottoms/3224851189/#in/set-72157612928904793/

2) Positive master quickly milled in a block of prototyping material:
http://www.flickr.com/photos/candy-bottoms/3224851197/#in/set-72157612928904793/
http://www.flickr.com/photos/candy-bottoms/3224851199/#in/set-72157612928904793/

3) Liquid silicone rubber poured into the masters to create flexible, multi-use negative molds:
http://www.flickr.com/photos/candy-bottoms/3251712414/#in/set-72157612928904793/

4) Liquid rigid plastic resin poured into the rubber negatives to cast final parts:
http://www.flickr.com/photos/candy-bottoms/3228983080/#in/set-72157612928904793/

While this may seem like a long route, it's actually much simpler than individually machining dozens of tiny parts directly from blocks of stock material of desired color and properties; that can be very wasteful and just hard to do right (workholding, etc).

ps - I'm still amazed you made gears and got them to work properly . . . did it work on the first try? Sometimes I can't even get screw holes to align on my first prototype

Nah, I cycled through several prototypes :-) About 40 hours of work total. There are some photos of earlier versions linked on the page:

1) First prototype - had several mechanical problems: http://lcamtuf.coredump.cx/robot/show_image.cgi/tinybotmk2-2.jpg

2) Second prototype - mechanically mostly OK, but decided to redo it to improve ground clearance and maximum turn angles: http://lcamtuf.coredump.cx/robot/show_image.cgi/IMG_1882.jpg

Cheers,
/mz

[Admin]

What brand liquid silicone rubber do you use? And does it hold up pretty well after many uses?

I occasionally use silicone rubber from smooth-on.com, and use a vacuum chamber to purify it (from micro-air bubbles) for increased strength.

I really like this technique . . . although I admit most of everything I do is a one-off part so not very useful for me at the moment . . .

[futmacl]

What brand liquid silicone rubber do you use? And does it hold up pretty well after many uses?

ShinEtsu KE-1310ST. It handles over 20 castings no sweat, never had a reason to do longer runs, but I expect it to survive 50-100. Condensation cure silicones or silicones with low tear strengths don't work nearly as well, though :-( I tried to capture much of this in that CNC guide, including a list of recommended resins (in section 7 - "Getting stock material").

I really like this technique . . . although I admit most of everything I do is a one-off part so not very useful for me at the moment . . .

So do I (well, it's useful to be able to cast parts several times over when things go wrong). But I found this workflow much faster and cheaper than ordering parts (few cents per part, about 6-9 hours from CAD model to a final part), and also easier than direct milling of complex shapes. Plus, you get to choose from more robust plastics... most people who do direct milling use HDPE (very flimsy) or acrylic (annoyingly brittle and suffers from stress cracking).

/mz

[dunk]

i have a question about your bot:
are you getting a reasonable turning circle with a solid back axle?

Turn radius is about 15 cm in this configuration, so it depends on your definition of "reasonable" :-) With a differential drive, I would need the robot to be a whole lot shorter to still be able to get at least some movement accuracy (but that sucks); or use a pivot / omni wheel, but that's difficult to instrument for traction sensing & dead reckoning.

interesting. 15cm turn radius is far better than i got with my early prototypes with a solid rear axle... that's pretty good.
the turning circle obviously depends largely on the distance between the rear wheels as well as the distance between the front and rear wheels.
even with the center of gravity quite far forward with ~20cm between my rear wheels and ~25cm between front and rear i would typically get a ~1m turning circle on full lock. not that good.

for me a single rear wheel performed far better.
the old LMbot would turn within 50cm.
http://sites.google.com/site/mrdunk2/lmbot

the thing i was always lacking was the ability to build a rear differential.
the required bevel gears would be an order of magnitude harder to manufacture but it looks like you have the tools and the patience....


dunk.

[futmacl]

even with the center of gravity quite far forward with ~20cm between my rear wheels and ~25cm between front and rear i would typically get a ~1m turning circle on full lock. not that good.

Axle distance is about 9 cm, front wheel pivot points are about 7 cm apart, wheels can turn about 30°. On carpeted surfaces, it's about 15 cm with very little slippage. On hard floors, probably closer to 25.

[Conscripted]

Nice robot,
I'd love to see it for sale!(i want one)

I'll second that.

Conscripted

[Hawaii00000]

Nice bot! That's some pretty hard-core robotics... Very nice.

[Gertlex]

@ futmacl

You are a sexy god of inhouse manufacturing.

@dunk

Single fixed rear wheel is just as good (geometrically) as a fixed axle for Ackermann geometry.  They only things that matter are the distance between front and rear wheels, how much the wheels can turn, and how well the turn angles match Ackermann angles.  Or that was my impression playing with that stuff for several months.

@ futmacl again

How much slop is there in the steering, due to the gears?

And geared steering => parallel steering, right?

So cool...

[futmacl]

How much slop is there in the steering, due to the gears?

Each gear has under 0.002 mm of backlash or so, so there is not much noticeable play; some angular play comes from the single-row bearings the wheels rotate in, but by the feel of it, it's not more than 1°.

You could also preload the whole thing with a spring or even just a rubber band if necessary, so this is not a big deal even with crappy gears probably.

And geared steering => parallel steering, right?

Yeah. I imagine you could make ideal Ackermann with non-circular gears if you really wanted to.

[dunk]

Single fixed rear wheel is just as good (geometrically) as a fixed axle for Ackermann geometry.  They only things that matter are the distance between front and rear wheels, how much the wheels can turn, and how well the turn angles match Ackermann angles.  Or that was my impression playing with that stuff for several months.

definitely not true.
that's why nearly all powered vehicles have a differential.

get some lego and build some free-wheeling chassis if you don't believe me.
start with one with a fixed back axle and a wheel base wider than it is long for an extreme example.


correctly matching the Ackermann angles would be nice but at this small scale the extra wear on tires of not doing so would be negligible.


dunk.

[Gertlex]

Single fixed rear wheel is just as good (geometrically) as a fixed axle for Ackermann geometry.  They only things that matter are the distance between front and rear wheels, how much the wheels can turn, and how well the turn angles match Ackermann angles.  Or that was my impression playing with that stuff for several months.

definitely not true.
that's why nearly all powered vehicles have a differential.

get some lego and build some free-wheeling chassis if you don't believe me.
start with one with a fixed back axle and a wheel base wider than it is long for an extreme example.

Edit: OMG...
http://lcamtuf.coredump.cx/robot/show_image.cgi/pmma_microbot.jpg
http://lcamtuf.coredump.cx/robot/
correctly matching the Ackermann angles would be nice but at this small scale the extra wear on tires of not doing so would be negligible.


dunk.

I used the word 'geometrically' for a reason.

In real applications, wheel slippage becomes a factor, and I can see single wheel rear noticeably being better in some cases.  (Don't have my legos with me for the summer :/ )