PDA

View Full Version : [Just For Fun] Glitch - a mini Twitch-like... thing



Xevel
04-03-2014, 04:39 PM
Hey,

I got some Dynamixel XL-320 today, and some Ollo parts to go with it... not wanting to wait for the moment when I will need them (they are supposed to probably integrate in my MW bot, but there is a sizable amount of development needed before I can assemble them), I thew a little bot together just to have some fun and discover the XL-320 and the OpenCM-9.04.

Here is the result of an afternoon of just assembling stuff at random until it looked like something that could move.

54855486

Let's call this thing Glitch, it does not deserve much more :p
It's based on TicTac6 / Twitch / TwitchMX, with some minor variations to make it possible to build with the XL-320. The "straight" legs are made with the batteries, and very close together (they sometimes touch at the bottom) to keep the weight distribution as low and centered as possible.
To that effect, I also oriented the vertical servos to have their axes as close as possible to the center, to avoid big swings. (losing some range of motion in the process).
The other legs are made to account for the flex of the material, and can surprisingly keep the little thing kind of upright ^^

The wheel at the top is their as a counterweight for the OpenCM, plus it swigs left and right... probably actually useless, but fun.

5484

Now, some coding should make it move.

I've tested the legs, they can easily support the weight, and despite the tremendous amount of play in... well, every joint, even fixed ones, it can lift one trio of leg completely from the ground and support itself on the others, as expected.

KevinO
04-03-2014, 05:55 PM
Pretty cool! So small. I guess I didn't realize their size.

jwatte
04-03-2014, 10:47 PM
Very neat!
Do post a video when it... wiggles.

Gertlex
04-04-2014, 12:20 PM
Dang, you beat me to building a Twitch-like with those new small servos. I am amused!

(I really should get around to doing the micro hobby servo one I was going to do...)

Xevel
04-04-2014, 06:16 PM
Hmm, Am I hallucinating or does the earlier versions of the XL-320 used the Dynamixel V1.0 protocol? Reading (http://www.robotcampus.fr/2014/03/05/beta-testing-dynamixel-xl-320/) some (http://aarondevlog.blogspot.fr/2014/03/chaining-dynamixel-xl-320-servos-with.html) blogs/posts and an un-linked page of Robotis support website (http://support.robotis.com/en/product/dynamixel/xl-series/xl-320.htm), it seems that at some point in time they did... The support website has been updated to reflect the use of the new protocol.

Mine use the Dynamixel v2.0 protocol, and while they use 1000000bps by default, the serial port of the DynamixelPro object has to be opendd with 3, not 1. Strange for a little thing like that to have it talk the same protocol as the Dynamixel Pro ^^

jwatte
04-04-2014, 10:48 PM
I thought they spoke DXL v1? In fact, the ones I have, do.

Also, they do run at 1 Mbps, but the baud rate setting is different from previous Dynamixels. There are only four baud rate values to choose from, 0 through 3, meaning 9600, 57600, 115200, or 1Mbps.

Perhaps the library that you're using for the OpenCM 9.04 is tuned for the XL-320. I'm using my own libraries, that I developed before the XL-320 were out, and configure it like it was previously done.

Xevel
04-04-2014, 11:05 PM
I use the DynamixelPro library of the OpenCM9.04, the examples with the regular Dxl 1.0 lib did not work (obviously).

Still, the fact that on the support website, the XL-320 now appears in the Dynamixel 2.0 section looks like a dead giveaway.

Xevel
04-06-2014, 02:15 AM
GlitchBot's first steps, on some shock absorbing silicon foam to avoid staying in place do to overabundant slipping ^^


http://youtu.be/_3oZYCOrNVk

It seems that the thing is aptly named, since there is an occasional glich in the ~10 lines of code consisting of a loop with some servo poses and some delays... As can be seen at the end of the video, sometimes there are servo commands that are not executed. Not sure if it's due to software or hardware, I'll have to investigate.

Anyway, the mechanical aspect works surprisingly well I think, for such a simple bot made with components that wobbly :)

I'm not even going to make this walk demo better, I'll probably directly jump to the full-fledged vector-based dynamic motion generation with smooth motion, curves, accelerations ramps, etc, that even they will be tested on this bot, will actually be for a bigger one I hinted at sometime ago in another thread, that will be autonomous and will participate in the French Robotics Cup, in 2 months.