Off Topic --> Artificial Neural Nets

I held a presentation on Artificial Neural Networks (ANNs) for a course the other day. I tried to make the slides as usefull as possible and decided to upload them here as well. They may be of use to someone with little knowledge of the topic looking for a simple introduction.

Anyway, here is the link: Presentation Artificial Neural Nets

Sewing the Stretch Sensor

I have been trying to figure out how to best connect the stretch sensors to fabric. What I did before was simply using relatively thick multi-strand cables and attaching the stretch sensor using pressure fitting. I would then just sew the multi-strand cable to the fabric.


However, the rigid plastic parts required for the pressure fitting do not especially appeal to me, they are clumsy and don’t work as well as I was hoping they would.
I was hoping to find a method of making the actual sensor more stable and making it easier to attach the sensor to clothing. I ended up sewing the sensor and cable directly to elastic bands. This works fairly well, especially as the maximum stretch of the elastic is less than that of the sensor, protecting the sensor from being overstretched. Well … enough talk, take a look at the pictures to see how this looks in practice

Ahh… and ignore all the extra threads and sloppy craftmanship. Next-time round this will look a lot prettyer. My sewing machine and me *are* (slowly) becoming friends…

ShadowCoat Mouse

I procrastinate productivly: I should have been doing other university stuff, instead I thought... "can I use my set-up as a mouse?"

Well, I gave it a shot. I'll coment more on it later.

Introducing: The ShadowCoat (Part 2)

As mentioned before,  I have changed my approach a bit, dropping the thermoplastics and going more towards smart clothing direction. I suggest you just take a look at the video. I'll post some pictures and explanations below.

What is most noteworthy in the video is that the rotation of the arm does the two sensor thing I've been talking about (cant wait to see that work on the elbow and shoulder as well). This greatly increased the accuracy of the reading. Also, the software now out-calibrates. This makes the suit ultra-usable. A friend just tried it on and it worked as effortlessly with her as it did with me.





I am still using the stretch sensor by Images. As you can't soldier plastic, I had to find other means of connecting them to cables:

  

I use multi-core cables (ahh, I know thats not the correct term, but its late...), so sewing over/through/on them is no problem

I am sewing them to the construction I prepared yesterday, you can see them in this post.

I was fooling around with the readings from the stretch sensors earlier and realized, that using two sensors which counteract each other would greatly benefit accuracy and amplitude of the signal. I tried it, and yep, it works :-)

From my ShadowCoat :-) the sensors connect to voltage dividers which in turn connect to the Arduino Mega...

Here is what the whole thing (and me) look  like at the moment. I'm quite happy, cause this worked better than expected :-)

Ad Thermoplast

A guy who goes by the name of cr0sh on the Arduino Forums, who has been giving me all kinds of interesting suggestions, wrote a comment on thermoplastics. I will just copy paste it over here, in case anyone who has been following what I am doing is interested.

Personally, while the stuff *is* amazing, I am quite happy to announce, that it no longer is the main structural component of my set up.

Anyway, here goes chr0sh's post:

I noticed you are moving to customized forms using thermoplastic (Turbocast from Streifeneder, Apolit from Minke Props).

Whether or not it will help you (though it might help others, so you might want to reference it on your blog?), such plastic is known by a couple of different names in English:

"PolyMorph"
"Friendly Plastic"
http://www.instamorph.com/
http://shapelock.com/
...among others - its chemical name is:

Polycaprolactone (PCL) - http://en.wikipedia.org/wiki/Polycaprolactone

One place I found that sells "bulk" (not sure if "to the public", though) is:

http://www.absorbables.com/standard.htm

They also have "sample kits". You can also find this plastic on Ebay quite often, typically sold as plastic for jewelry making; it comes in a wide variety of colors (including metallics). Standard forms I have seen (before seeing your samples), have been a flat "stick" form, and pellets.

In the US, I have seen it sold in craft stores, and the friendly plastic is available from Target stores.

By the way, cr0sh has a website: phoenixgarage.org check it out :-)

Introducing: The ShadowCoat

I havent updated here in a while. I have sort of important news though (well, in a way)

I have decided to call my input device ShadowCoat because well, it copies your moves, like a shadow... and your can wear it, like a coat :-)

Oh, and for those of you who are thinking... "this thing is wearable, but no where near as comfortable as a coat..." ... well, I got out my sewing machine... I am not really ready to present what I am sewing, but I thought I might as well upload some pictures in the meantime :-)

 

I also built myself a new ArmBot: ArmBot II :-D

... this one is not as pretty (and thats a euphemism), *but* it actually has all the degrees of freedom I am measuring. This is a huge help - having something physical in front of me to controll makes things much clearer to me.

Sending Data over the Network --> First Success

So I figured out that Processing has a network library. Its a step in the right direction. Raises the question weather this can be used for sending data via the Internet. Probably not. Actually, yes it can :-) But hey. One step at a time

Enjoy the clip:

Some Pictures of new Prototype in Action

Connecting Thermoplast to Sensors

(I will comment on these later. Right now I just want to upload these so I can fool around with some new stuff.)

Turbocast: Low Temperature Thermoplast (Streifeneder)

A couple of days later I received a parcel from Streifeneder with some of the 'Turbocast' which they produce.

 

Turbocast was much more difficult to mold. It stuck glued to my tools. It glued to the hair on my hand, which was sort of painful ... (I ended up shaving my hand before continuing experimenting with Turbocast). Somehow this material just felt more stubborn than Apolit.

However, eventually I got used to the material and was able to get some decent results out of it. More on that topic soon.

Apolit: Low Temperature Thermoplast (Minke Props)

A couple of days ago a received a sample of Apolit, a low temperature Thermoplast by Minke Props.

First off, I have to say that I love this stuff. Where it not rather expensive, I would buy a ton of it...

It was quite easy to mold, I did not burn myself, nor did it pull out any hair after having hardened on my hand. If you need material for instant prototyping this is very interesting stuff.

Take a look at how my first encounter with Thermoplast went.

If you toss it into boiling water it turns soft and transparent in under a minute. I assume you could also bake it in the oven, though it would definitely need a much higher temperature. I was a bit worried that this material might melt if left in a car or something. However I think that would require very, very high temperatures (not saying these are impossible) as air is not all that good in transporting heat. I think its really the fact that water 'transports heat' so well which makes this work so effortlessly.

If you don't like the way its molded, just throw it back into the water. It will go back to its original (flat) shape and become moldable again.

Cutting it is no big deal, however if you use thicker material that could be a pain. I figured this out later, but obviously if you cut it while its still soft. A knife or scissors would go through it like butter.

I also tried sanding it, which did not work very will. Sanding it feels like rubbing an eraser on paper. Carving it with a sharp knife works better. Sawing it should work fine.

I added some self adhesive Velcro. This works sort of fine, but if you want something really permanent, you may want to additionally secure it with screws or something. 

By the way. I hope to soon receive a ton of sensors by images. This 'wristband' will then be used for mounting the stretch sensors for measuring arm rotation. More on that in a later post thought.

Getting better aquainted with the Stretch Sensors

As mentioned in a previous post, I now have a stretch sensor by Images in my possession. Though I was first a bit skeptical I am beginning to see more and more ways of using this. I sort of had a hunch that the relationship between stretch and resistance might not be linear (yeah admitted, nothing strange in that, I would be more surprised *where* there a linear relationship.)

I am a visual person and I need to see things to understand them, so I made these graphs:

Stretched to 125 % of original length.

Stretched to 150 %

Stretched to 175 %

Stretched to 112.5 %, 125 %, 137.5 %, 150 %, 162,5 %, and 175%

Stretched with incrementions of 6.25 %

 

Due to the fact that I am lazy the graphs might be a bit confusing to read. The x-axis represents time and the white vertical lines represent seconds (I just double checked my math and the white lines are more like half seconds - maybe I will redo these if I get around to it. However you still get the generally picture). The y-axis measure the change in resistance. The light gray area is the resistance of the sensor. I did not add units, as the interesting part is the relative change and that is all I measured. 

The fact that the signal is non-linear can, with some tinkering, be used as a bonus. The resolution in the first couple percent of stretch is quite high - so if I where to use two stretch sensors simultaneously but in opposite phase (if one is loose, the other one is stretched) I should be able to maximize the resolution available :-)

Feedback Circuit - Signal Flow

Had a productive meeting with the other PEERS / Research Based Learning people. I finally feel like I have an overview of all the different areas I have to tackle for this. I like making graphs to sort my ideas: This roughly represents the feedback loop which I hope to create.

Putting the Pieces together

Well… of course I did not sit down to do my readings for UCM as I had planned. Instead I made another video demonstrating how flex and stretch sensors can be combined. Together with the potentiometers I basically have all the sensors I need now. I hope I will be receiving Thermoplast some time this week to build a stable frame for all this tech which is, as of now, sort of floating around..

Ok. Here goes the video:

Images Stretch & Flex sensor

(scroll down a bit to go directly to the videos.)

As I said, I received two sensors by images the other day. A flex sensors and a stretch sensor (find documentation here and here. Unlike a potentiometer these do not come with a voltage divider to simply measure the resistance, so I had to go about constructing my own. Basically what I did was put a resister between ground and the sensor and attached a third cable to the sensor right before where I attached the resister. (The same way as you would do it for a photo-resistor or any other resistive sensor i guess...)

Here are some pictures of the first readings I took with the flex sensor:

straight:

and bent:

I then went about to see if I could actually use them as I hoped. (The flex sensor for measuring the hands clasp and the stretch sensor for measuring wrist rotation)

Here's an image of my flex-sensor-glove setup:

And here are videos of both flex and stretch sensors in action:






I am a bit worried that the signal from the stretch sensor might be too weak, but I believe I can improve this when I actually go about constructing the whole thing (I might not be using the optimal resistor in the voltage divider and have to experiment with location of the sensor etc.)

The flex sensor/glove combo works so well that I've had everyone who comes close to my room wear it and try it out for themselves :-D