PDA

View Full Version : TwitchMX - The Mech of Many Feet



Gertlex
01-16-2013, 10:14 PM
I'll just start this one off similar to Twitch's thread -- Look, Ma! No feet!

44624463

The mechanical stuff is not even bolted together. I've got lots of turret work yet to do, and leg mounting to re-design. Oh ya, and battery and power switch location are completely TBD.

The electronics are two Wixels (256 bytes of RAM each!!), both on the robot side, and an xBee will go to the Arbotix Commander. If I have time, I'll have a second software configuration where a third Wixel on the Arbotix Commander is used... However since xBee range > Wixel range, the former gets priority. I'll be using a Pololu motor driver board, and a probably a micro servo hopper agitator, as well.

Software is just a matter of doing and testing. Most of the stuff I've not done yet (gun control) I can almost directly copy from Numa's code.

jwatte
01-17-2013, 09:06 PM
See you in April! :-)

Gertlex
01-20-2013, 08:51 PM
Did some fabrication today. Now that the body is done, I have a better sense of where things go, and so now back to CAD work. Also time for a McMaster order since I need longer 2.5mm bolts for the MX servos...

4474

The turret mounting is two 3D printed parts "welded" together with ABS filament in a dremel at 10 or 20 kRPM. I tacked it in place while the bot was assembled, then sealed the entire perimeter once I removed the parts. There are also a few bolts on the underside that are also holding the pieces together.
4476

And this was nice improvisation to make dealing with the nuts a bit easier...
4475

jwatte
01-20-2013, 10:00 PM
The springs to hold the nuts in place are brilliant! I've been using Scotch tape, but that's not something I can leave while running, and I forget to re-tape it when disassembling, and the nuts fall out.

Also: Are the black horizontal bars also 3D printed?

Th232
01-20-2013, 10:10 PM
Also time for a McMaster order since I need longer 2.5mm bolts for the MX servos...


Haha, story of my life. I dread to know how many hundreds of screws I have at the moment. I like how you've done the turret as well.

tician
01-20-2013, 10:18 PM
The turret mounting is two 3D printed parts "welded" together with ABS filament in a dremel at 10 or 20 kRPM. I tacked it in place while the bot was assembled, then sealed the entire perimeter once I removed the parts.I've never seen that particular method employed before, though there is something similar called spin welding - where a metal tool spins fast enough in direct contact with the two pieces to soften and blend them together without the really damaging/weakening high heat of 'normal' welding (I seem to recall it being used on some of the later generation aluminum tanks for the space shuttle - hopefully I have fulfilled my daily quota of random crap nobody really needs to hear).


And this was nice improvisation to make dealing with the nuts a bit easier...
Very nice. Reassembling a DARwIn-OP can be a bit challenging when the nuts don't want to stay in their spots on the mounting flange. Should work to hold them in place long enough to get the bolts started, then pop the springs out later to prevent weight distribution problems. I've had to use index cards as filler between the servo case and plastic frames when assembling AX frames to keep the nuts from falling out (not nice knowing it is there, but it is hidden and works well enough when you don't have any extra black plastic wire routing bands from a premium kit).

Gertlex
01-20-2013, 10:57 PM
I've never seen that particular method employed before, though there is something similar called spin welding - where a metal tool spins fast enough in direct contact with the two pieces to soften and blend them together without the really damaging/weakening high heat of 'normal' welding (I seem to recall it being used on some of the later generation aluminum tanks for the space shuttle - hopefully I have fulfilled my daily quota of random crap nobody really needs to hear).

In the case of metals, it's called friction stir welding - and yeah, softening without melting is the key (I re-read the wiki article a couple days ago ;) ). Doing this with plastic was something that showed up on Hack-A-Day a couple of weeks ago, and it works great with easy-to-melt ABS plastic. I was going to buy some 1/8" polycarb rod on McMaster to try, but that comes in 8ft lengths....

I'm in the process of dis/re-assembly, putting the wires on the MX servos. A true PITA.

Edit: Hah. I was sure I ID'd all three servos when I got them, but I only did the middle one. So now I need to remove a wire again while I set the outside servos' IDs... Joy :)

DresnerRobotics
01-21-2013, 08:09 PM
All of you complaining about nuts & bolts with the dynamixels... you shush your mouths. I'm not sure anyone outside of Robotis has assembled/disassembled as many Dynamixel based robots as I have in the last 5 years. Ugh.

Gertlex
01-21-2013, 11:58 PM
So the path to minimizing communal misery is to have you assemble our stuff, too? :D

byi
01-22-2013, 08:49 AM
nice welding technique. i might use that. im very jealous of the 3d printer. i wish i had the budget/time for one.

Gertlex
02-17-2013, 11:17 PM
Progress has been lots of small things recently. Next weekend I should have legs in place.

Most recently, I've got the turret linkages assembled. Next up is to get the turret shell/box designed, and prototyped with cardboard... gotta verify placement of all the electronics up to. Similarly, I need to prototype the battery case, but that should be easier to do.

4527
I'm using a lot of that brass rod... 36 shaft collars and nearly 5ft of brass rod on the final bot :eek:

Software progress has also happened. However, testing is pending my mounting the electronics, so I need to make that a priority.

jwatte
02-18-2013, 03:08 AM
Is that an AX-12 on top? I'm trying to figure out how it's actually linked into your turret. Does the black gear drive the entire carriage forward/back at the bottom, and thus aiming up/down?

tician
02-18-2013, 03:52 AM
If I am thinking correctly, there is a black printed gear attached to the left AX-12. That meshes with the printed gear on the vertical frame connecting the gun mount to the rest of the bot (gear is fixed to frame, with a swivel directly above it connecting the fixed frame/gear to the gun mount). As the pan servo rotates, it causes the entire gun mount to swivel on the frame. The tilt servo is attached to the double wide dynamixel frame usually used in the hip and foot of a bioloid humanoid. I think it should permit less play in a shorter form factor than mounting directly to the servo horns with (like on the phantomx pincher arm) or without external thrust bearings (kinect pan-tilt on maxwell and darsha), and kinda wish I had thought of it.

Gertlex
02-18-2013, 09:28 AM
Jwatte, I'm using the exact same geometry for linkages as I describe in this post for Numa: http://forums.trossenrobotics.com/showthread.php?4929-My-4DoF-Quad-Mech-Numa&p=50540#post50540

You can see the turret attachments more clearly in the first post of this thread.

Tician is basically correct, except the gears are from the internals of the Crosman airsoft guns :D (My printer has nowhere near the needed quality for gears) In this case, I saved about half an inch in height. I still had to have the printed riser so that the leg linkages don't hit the turret, though.

Gertlex
02-24-2013, 08:52 PM
Spent the weekend CAD'ing and laser-ing at Trossen.

Prior to that I did some work on the adjustable end legs. The below mechanism slides, with a bolt that I can turn to get about 1/4" of height adjustment. I then have several holes, so I can move the (yellow 3D printed) slider over a wide range of foot heights. These pics aren't the final product - I counterbore such that the bolt head is within the 3D printed plastic, and then plates bolted above (to which the legs mount in the second pic) hold the bolt head in place.
4532

I went with the "laser-cut box" approach for TwitchMX's canopy. Here's an initial shot.
4531
I still need to attach the top and back panels, and mount the electronics. Gotta do some 3D printed parts first, though. I'll have better pics eventually ;)

The battery carriage is also new. It's hair pins and 0.03" polycarbonate...

kamondelious
02-24-2013, 11:06 PM
Looking good Sir!

Can't wait to see this bad boy walking.

Gertlex
03-02-2013, 03:56 PM
Everything but the guns (and score panels/LEDs) is wired up.

Other than the power switch being built into one of the legs, all the electronics are in the turret and easily accessible.

Rear view. A single hairpin goes through two holes in the top, and clips around a bolt.
4536
Remove the hairpin and...

Wires! Back and top fold out. Rubber band keeps stress of wires when the back is open, and I stash the hairpin on the bottom right so I don't lose it...
4535


Side view. The single gun slides onto the brass rods sticking out on other side. Gotta trim this side's rods down.
4537

Turret rotated at an angle. Note connector for cam's auxiliary atenna. Blue LED is the battery monitor - the plastic is rather translucent and one side is fairly reflective.
4538

Gertlex
03-02-2013, 05:18 PM
And one more pic.
4539

Gotta tuck that battery cable in a bit, I think :)

jwatte
03-02-2013, 07:31 PM
Shiny!

Is that gray plastic acrylic?

tician
03-02-2013, 07:38 PM
I wouldn't expect plexiglass to survive very long against an airsoft barrage. I seem to remember someone making a protector for their camera from plexi and having it shatter horribly on one of the first hits.

byi
03-02-2013, 09:47 PM
Wow. Looks really great. Makes mine look like it was pulled out of a trash compactor.

jwatte
03-02-2013, 09:48 PM
I wouldn't expect plexiglass to survive very long against an airsoft barrage.

That was my feeling as well, hence why I asked. But if the plastic is reflective on one side, and matte on the other, it probably isn't acrylic, because that's typically reflective on both sides.
The question, then, is what is this plastic that cuts nicely on laser cutters :-) (I assume it doesn't have chlorine in it like ABS, although perhaps Trossen has a heavy duty gas cleaning setup for their cutter?)

Gertlex
03-02-2013, 10:42 PM
It's a plexiglass, as the sheet the parts were cut from was labeled. Tyberius swears by it. One side is indeed textured, which is why I have lots of non-reflective outer surfaces. The lasers at Trossen do have a ventilation system.

As for cracking, I'm not too worried about catastrophic issues. Some visual cracks will add to the appeal :). Most of the turret surface will be covered by score panels, too.


I wouldn't expect plexiglass to survive very long against an airsoft barrage. I seem to remember someone making a protector for their camera from plexi and having it shatter horribly on one of the first hits.
My camera has a square of .03" polycarbonate protecting the lens. You can see it reflecting stuff in the last pic.

Gertlex
03-10-2013, 09:00 PM
http://www.youtube.com/watch?v=kAxg_pC-6jU

Lots of tweaking to be done, especially with turning gait.

Vid coloring is horrible... yup. I used my modified MWCam.py to get TwitchMX's view, and then OpenCV (also in Python) to get my Surface Pro's webcam stream too. In the latter case, it may be that the camera depends on software to correct the color. [/trivia]

byi
03-10-2013, 09:06 PM
Wow. Really stable and fairly fast. If you say there is room for improvement, I am sincerely scared.

kamondelious
03-10-2013, 09:58 PM
Great job Gertlex!

DresnerRobotics
03-11-2013, 03:12 PM
FYI- we're ruling this a quadruped for this year's competition purposes. Given the current rules, feet & leg structure aren't clearly defined. This may change with a rule revision next year, but for now it's going to be classified as a quadruped pending further discussion.

ArduTank
03-11-2013, 07:26 PM
Gert, one recommendation I have is: try to get the turret to smooth out the left - right movement when walking. It'll help when aiming on the move.

Gertlex
03-12-2013, 12:12 PM
Gert, one recommendation I have is: try to get the turret to smooth out the left - right movement when walking. It'll help when aiming on the move.
I'm a step ahead of you there :) I've had initial code/gait (https://github.com/erelson/wixel-sdk/commit/26bf4cf8162ce94fde06984c80e539c4e22e264e)additions for this for a few weeks, but not enabled/tested it yet.

What I really ought to do is figure out the full kinematics and from that, optimizing gaits such that (a) walking is as smooth as possible, and (b) I know exactly what positions to send the pan servo so that the turret motion is smooth while walking.

ArduTank
03-12-2013, 03:11 PM
Nice. I, unfortunately, still have yet to get M-O new legs sections so that I can get him walking. I just don't have the materials right now.

Gertlex
03-13-2013, 09:51 PM
Despite all my talk of fancy math solutions being ideal, the simple turret panning gait while walking is very effective.

Now to figure out why the D-Link camera stopped connecting to my router...

Gertlex
04-06-2013, 01:17 PM
D-Link camera likes to reset itself. Ahh well, easy enough to re-setup.

I got the VNH5019 motor driver to be driven by the Wixel. Had to write that library myself. Took a couple weeks of reading code and reading the manual for the TI chip on the Wixel. Definitely have a better understanding of some of that low level stuff now.

Also have a laser wired up as well. Definitely love having a 3D printer to make mounting cylindrical things easy.
4614

TwitchMX is fully functional. Working on various tweaks now. I'll do my prequal vid next weekend most likely.

Gertlex
04-14-2013, 02:58 PM
Prequal vid:


http://www.youtube.com/watch?v=J_1cnK3u-FU&feature=youtu.be


Here is TwitchMX's prequal. Shot with my spare D-Link camera attached to the chain of a hanging light fixture :D

6 targets... Gun jammed while I was going after the last one. First jam in several weeks, bah.

And of course Numa makes an appearance and gets shot in the face. (I used empty cupcake liner containers with post it notes for visibility.)

ArduTank
04-14-2013, 06:21 PM
Nice. I see you got the turret left/right motion smoothed out when walking.

jwatte
04-14-2013, 07:43 PM
That is pretty darn nimble! Maybe I'll crash-course something like it to get a 'bot up and walking...

Honestly, that looks way too stable! If you win this year, maybe "points of contact with the ground" or "static stability even when moving" will be the next definition of qualifying bots :-)

Gertlex
04-15-2013, 09:40 AM
Gun unjammer via solenoid and 3D printer because I could...

4632

(The 3D printed part isn't actually visible)

Gertlex
04-16-2013, 12:23 PM
I mounted the lighting board. I was then all set to declare that TwitchMX was done. Hit the solenoid button on my Commander, and bam, self-triggering the scoring system. d'oh. Testing last night with wires everywhere suggests that I can fix it by powering the scoring system with a separate 5 V switching regulator (even though the scoring system and relay are already using separate 5 V switching regulators...), so I'll make those changes soon. Worst case, I sacrifice a hitpoint to unjam the gun occasionally :P

byi
04-16-2013, 12:42 PM
I just run a servo continuously to keep flow going. just a thought if you dont want to trigger your scoring system.

jwatte
04-16-2013, 01:36 PM
My solution is very small gear motors with an agitator:

http://forums.trossenrobotics.com/gallery/files/7/4/3/2/bb-magazine-progress_original.jpg

Anyway, if you want super clean power, then a 7V switching regulator followed by a large capacitor and a fast 5V linear regulator is the gold standard. That could get you audiophile quality isolation (> 100 dB rejection of input spikes.)

ArduTank
04-16-2013, 05:03 PM
Lol. M-O's body is large enough to not get near any scoring system wires.

Gertlex
04-16-2013, 06:12 PM
Lol. M-O's body is large enough to not get near any scoring system wires.

Testing showed that proximity of the wires has absolutely nothing to do with triggering my scoring system.

Also I already noted that I have a solution (which I just need time in which to implement).

ArduTank
04-17-2013, 06:09 PM
True.