Modified Humanoid Type A with tilt/pan inside Premium shell

[garycmartin]

I've been tinkering with the standard Humanoid Type A for a while now, but since getting a HaViMo2 and two extra servos a couple of months back for a camera pan/tilt set-up I've finally decided to have a fairly large rebuild of the Type A upper body. The goal was to try and keep the Premium clear plastic shells, useful protection for the camera with all the inevitable face first head plants. Here's the flickr set I've been uploading to:

http://flic.kr/s/aHsjuospRC

Seems to have worked out pretty well so far with only a couple of physical hacks (needed to make a notch in one bracket, and the inside front ridge of the face visor area to be filled down). I'm only just starting to rework actual motions/tasks for the new topology, will try and post a movie or two as I progress.

--Gary

[i-Bot]

Interesting you changed the arm. DARwin also has the elbow rotated as you have.

I too am trying to get a neat head pan/tilt into comprehensive and premium, but want to maintain the DARwin joint relationship.

Waiting to see how you get on with changing motions

[garycmartin]

Interesting you changed the arm. DARwin also has the elbow rotated as you have.

I too am trying to get a neat head pan/tilt into comprehensive and premium, but want to maintain the DARwin joint relationship.

Waiting to see how you get on with changing motions

FWIW: with a little more tweaking to align the re-build's servo horns in more friendly orientations and some adjustments to the brackets, most of the original Type A motions work without change. Here's some images of the visible changes:

http://flic.kr/p/9yaW3o
http://flic.kr/p/9y7Xrz
http://flic.kr/p/9y7Xhr

I've primarily been testing the walking/turning motions which all seem fine, the getting up when lying on its back motion does need a rework, but many others are just fine. Balance is a little more unsteady than a default Humanoid A, seems the head tilt servo is the main issue as removing it make walking just as stable as before – nothing major, I'm sure some tweaking of the default motions will help.

--Gary

P.S. Following your Darwin-OP posts with interest.

[PedroR]

Hi Gary

very impressive work with the PREMIUM kit. It does look a lot more robust I must add and great use of the space inside the chest and also a very ingenious solution to maintain the head and still do pan/tilt!

I was just wondering where you mounted the HaViMo as I don't seem to be able to find in it in some of the pics.

Thanks
Pedro.

[garycmartin]

Hi PerdoR,

very impressive work with the PREMIUM kit. It does look a lot more robust I must add and great use of the space inside the chest and also a very ingenious solution to maintain the head and still do pan/tilt!

I was just wondering where you mounted the HaViMo as I don't seem to be able to find in it in some of the pics..

Thanks, have a look at http://flic.kr/p/9xhetJ as in the last few images uploaded I had already removed the HaViMo for easier external re-testing. I've added a note to the image - the HaViMo2 is mounted 90deg rotated, with the camera looking out just to one side of the bridge of the nose, the plastic there is nice and flat so minimal visual distortion. All the head cabling runs down the back of the neck so it's nice and clean (one Dynamixel bus cable, and the DMS lead to an analogue port). Putting the DMS inside the shell does impact the quality of the data it reports, but still usable for autonomous walk type behaviour (I might try mounting it further forward and see if there is a sweet spot somewhere for best results).


Added 2 DOF spine by Gary C Martin, on Flickr

Regards,
--Gary

[PedroR]

I see... very good design indeed!

This is likelly the best looking Bioloid PREMIUM with a HaViMo fitted that I've seen

Once again, congrats on the nice work!

[sharpyuk]

This looks great!

How did you get the AX12 in the head? Did you have to file down the head opening at the bottom to make it fit?

I'm trying something similar but it doesn't appear to fit without modification..

[garycmartin]

This looks great!
How did you get the AX12 in the head? Did you have to file down the head opening at the bottom to make it fit?
I'm trying something similar but it doesn't appear to fit without modification..

Yep, you need to file the front lip of the mask down otherwise it collides with the edges of the servo casing. The Havimo2 and DMS can both fit inside. I ended up placing the Havimo2 camera peering through the forehead area, with the DMS above and behind it, its accuracy somewhat reduced but still usable (I assume due to internal reflections).

--Gary

[sharpyuk]

Yep, you need to file the front lip of the mask down otherwise it collides with the edges of the servo casing. The Havimo2 and DMS can both fit inside. I ended up placing the Havimo2 camera peering through the forehead area, with the DMS above and behind it, its accuracy somewhat reduced but still usable (I assume due to internal reflections).

--Gary

Excellent, I was going to ask how well does the DMS module work inside the head.. I had a feeling it wouldn't be as good placed in the head.

Cheers

[sharpyuk]

I notice in this picture, the robot is very tall, does it work well this big? I would think it would be a long way to fall..



http://www.flickr.com/photos/50394334@N ... 464170730/

Do you have any videos of your robot walking with the longer legs and hip replacement?

Cheers

[garycmartin]

I notice in this picture, the robot is very tall, does it work well this big? I would think it would be a long way to fall..

Yea, it's at about 57cm tall and close to the servo limit. I'm using a fairly springy servo setup for the upper body so if it does fall it distributes the impact somewhat, but I'm still fairly careful and I'm their to catch it or have a cushion or two for it to land on.

http://www.flickr.com/photos/50394334@N ... 464170730/

Do you have any videos of your robot walking with the longer legs and hip replacement?
Cheers

A fair request With a biped this structurally unstable/weak I've been working with realtime servo balance reflexes, and now avoid using any pre-generated Motion animations. I've recently switched over to using Python so I can hack and test in a more comfortable dev environment than trying to use Task with the micro-controller, but I did have a passible if slow leg to leg shuffle and balance working on the micro controller a couple of weeks back. I'll track down the tsk file and make a quick video.

--Gary

[sharpyuk]

Right ok, I've not seen Python.. I'm trying to code mine using the embedded C. Just getting used to WinAVR as I've always programmed PIC microcontrollers in the past.

Essentially that's what I would like as my end goal, but it's going to take me a LONG time to get that working!

[garycmartin]

I notice in this picture, the robot is very tall, does it work well this big? I would think it would be a long way to fall..



http://www.flickr.com/photos/50394334@N ... 464170730/

Do you have any videos of your robot walking with the longer legs and hip replacement?

Cheers

Here you go, finally got this recorded and uploaded today:

http://youtu.be/MuxLWeDFxiQ

--Gary