Tuesday, December 29, 2015

Artificial Muscles Actuating Things

They're still not terribly fast, and they don't move terribly far, but I've arrived at the crucial step of getting artificial muscles to actuate something more than just a weight hanging from a string. First up, we have a heterochiral muscle (the type that expand when heated) flexing a piece of paper. Each coil of the muscle is sewn to the paper on one side with a loop of thread, so the coils expand on one side and are constrained on the other, causing the paper to bend. This is almost a no-load movement, and strikes me as being most useful for something decorative, such as an artificial plant. You might notice that the cooling cycle is almost as quick as the heating cycle. I attribute that to chilly ambient temperatures in the upstairs laboratory.

Next we have a homochiral (contracting) muscle, rotating a piece of cardboard on a hinge. The opposing force of the rubber band on the opposite side pulls the cardboard back into its original position during the cooling cycle. The homochiral muscle featured in the video has been annealed with some space between the coils when at rest, so it doesn't have to be put under tension in order to have working room.

Both of these muscles are drawing about 1 A of current. They are made of Trilene Big Game fishing line, test strength 50 lb., diameter 711 um, with a heating element of 10/46 copper litz wire. The homochiral muscle has two strands wired in parallel. The heterochiral muscle was coiled on a rod of 3/16” diameter, while the homochiral muscle was coiled on a 1/8” rod.

Annealing muscles with built-in coil spacing

I determined in some previousexperiments that trying to spread the coils of a homochiral muscle when one first coils it on the rod is a bad idea. At this point there is a great deal of tension on the line – it wants to uncoil itself – and the coils don't cooperate very well. Coiling the muscle on a threaded rod is a possibility that I haven't tried yet; the spacing of the threads would limit the coil spacings one could achieve.

I started looking at ways to adjust the coil spacing after the initial annealing. First, I tried putting the muscle under tension (with no rod in the center) and running a high current through the heating wire, hoping to anneal it into its new shape. This method gives the most even coil spacing one could ask for, but the amount of heat applied to the muscle was only sufficient to “soft-set” it. I noticed that as it sat around for a few days, the coils slowly returned to their original close-packed configuration. When I tried annealing a muscle under tension at full heat in the oven, without a supporting rod in the center, the coils just went flat.

In the end, the best method I found was to put the muscle through its first annealing phase, manually spread the coils on the rod, then anneal it a second time. It's a little tedious – friction holds the coils against the rod, so you have to slide each one into the right position with your fingernail to get the spacing even – but it seems to work.

A close-up photo of the homochiral muscle with spread coils

An aside about plastic springs

In addition to artificial muscles, you can make simple passive springs by coiling nylon monofilament around a rod and annealing it (without primary twisting or a heating wire). The spring constant is determined by the thickness of the filament (larger diameters yield larger constants) and the size of the rod (smaller diameters yield larger constants). I got rather excited about this a few months ago, thinking I'd never need to buy a spring again. The problem is, these plastic springs don't necessarily hold up well.

If you follow any of my social media feeds, you might remember when I posted this spider leg video. There's a plastic spring at each joint, made from the 533 um Zebcom Omniflex line, and I'm actuating them by pulling the tendons with my fingers:

I took that video the day I finished building the leg. A few days later, the leg was in sorry shape, merely because I had allowed my house to heat up in the afternoons. This was sufficient to make the springs relax a good deal, so that I had to shorten them to get the same degree of tension I had before.

Naturally, this leaves me in some concern about the muscles as well. How might they be affected by high ambient temperatures? I haven't done any tests in which I compared a muscle's performance across many sessions of operation, with temperature spikes in between.

Blog news

Comments now require moderator approval, because spammers have been really junking up the place. Sorry.

In the new year, I think I'm going to try to build an “artificial muscle summary” page featuring everything I've learned, for the benefit of others who want to experiment. Once that's done, I may set muscles aside for a while. There are soooo many other things I want to work on.

Have a most excellent New Year!
-- Jenny

Monday, August 31, 2015

DIY Nylon Muscles VII

IT'S A NEW MUSCLE POST, EVERYONE! Sorry, I got distracted by other projects and life in general. One note before I get into my most recent work: I am now collecting other blogs/sites that deal with nylon muscles in the right-hand sidebar. If you run or know of a site that isn't listed, please tip me off so I can include it.

There were several things I got very tired of while experimenting on these muscles, and one of them was trying to measure deflections on the order of millimeters by squinting at a ruler mounted beside the muscle. Dangling everything from the edge of the dining table or ottoman was a bit awkward too. So I built myself a muscle test rig out of scrap wood. It provides a place to suspend an actuator and a weight and converts the linear motion of the actuator into the angular motion of a long needle. Small movements at one end of the needle are amplified at the other end, making them easier to see and measure.

The first version of the test rig just had a needle which pivoted on a piece of stiff wire driven through the board behind it. The pivot point was located very close to one end of the needle, and on the short end there were too loops of wire attached to the needle: one to connect the muscle, and the other to connect the weight. This arrangement left some things to be desired. For one thing, I could never get the wire perfectly straight, or constrain the needle so that it would lie flush with the graduated backdrop. That meant the needle's rotation was not planar, or it would stick as it turned, etc. On top of that, the motion of the short end of the needle didn't leave the muscle free to move straight up and down. Near the needle's zero point, the motion of the muscle is approximately vertical … but as the needle continues to rotate, its end begins moving more and more in a horizontal direction, changing the mechanical advantage the muscle has and introducing complications that I would rather not deal with.

 Left: Muscle suspending weight on deflection test rig, version 1.  Right: close-up of the reel on version 2.

Wanting something better, I replaced this lever system with a reel. A screw through the center of the reel provides it with a rotary axle. Two strings are tied to holes in the reel and wrap around it so that unwinding one string winds the other. When both strings are put under tension, they remain perpendicular to the side of the reel, regardless of the reel's angular position. I connect one string to the muscle and hang the weight from the other string, and the muscle is forced to lift the weight when it contracts. The needle is a piece of thin aluminum tubing inserted into a hole in one side of the reel; I can “zero” it by adjusting the length of string between the muscle and the reel. (I knotted multiple loops in the string, and a piece of flexible wire between the muscle end and one of the loops is helpful for getting things just right.)

For testing muscles that only contract over a short distance, this thing is amazing. No more staring at the muscle and thinking, Huh, is it doing something? I'm not quite sure. When the muscle starts moving, I get an obvious deflection out of the needle. The biggest remaining issue is that there's enough friction and/or elasticity in the system that there isn't a well-defined zero point for the needle. For a given muscle-and-weight setup hanging passively (muscle is turned off), there's a fairly wide angular range within which I can position the needle and have it remain stable. When I take a muscle through a heat-cool cycle, the needle generally doesn't return to its original position at the end of the cool cycle. How much of this is due to the muscle stretching out and how much is just the equipment, I unfortunately can't say.

The other little quality-of-life improvement I attempted for this round of muscle experiments has to do with ease of manufacturing. I was sick of going through the effort of making a muscle, only to have the fragile heating wire snap at the last minute. When that happens, the nylon can't be returned to its pristine state, and now the wire is too short – so often I would be forced to throw everything away and start over. Putting extra slack in the wire could result in bunching and loose wire coils, promoting uneven and inefficient heating of the muscle. So I tried a couple of different methods to relieve strain on the wire and keep it unified with the nylon.

Coiling a muscle with tape tags.
 For Method 1, I tried attaching the wire to the nylon at intervals of a couple inches, using little tags of adhesive tape. These can be removed after annealing by sliding a straight pin in next to the nylon and pulling outward to separate the two sides of the tape tag. Method 2 was a little more wild: I tacked the wire to the nylon by coating both with a thin layer of silicone caulk. Messy as it sounds, I found that the best way to apply this was to stroke it on with my fingers. It cleans up just fine with some mineral spirits (paint thinner).

In the end, I'm not sure if either trick helped a lot. For each method, I made four muscles and lost one out of the four (due to snapped wire). That's not horrible, but certainly not great either. I did seem to get nice even coiling of the wire around the nylon.

Besides trying out these manufacturing tricks, the principal experiment for this month involved rod-coiled muscles with spread coils. I had previously noted that the muscles with a smaller coil diameter could lift more weight, but had difficulty achieving a good contraction distance because their coils were already so tightly packed. I thought that coiling them around the rod with some spacing between the coils might improve that situation.

Actually doing this turned out to be harder than I expected. Homochiral muscles naturally form close-packed secondary windings. You have to fight the muscle to get it to lie on the rod any other way – and the small-diameter ones fight pretty hard. What you see in the photo is the best I could do. The mandrel diameter used for all of these is ~1 mm (large size paperclip wire). One of each type has a silicone coating, and one doesn't. The close-packed “controls” are on the top, and the ones with spread coils are on the bottom.

Yeeccch. Those look terrible. But I decided to see if they would work anyway. I used a current of ~220 mA and ran a bunch of tests with different weights. All the muscles were allowed to heat for at least 4 minutes and cool for at least 9 minutes, with the idea that this would be sufficient time for them to reach “steady state.” Results are given in terms of the needle displacement in degrees, and represent the maximum distance the needle moved from whatever its initial position was. An entry of “failure” in the table means that the muscle started to stretch under the load when heated, i.e. the needle displacement was negative. None of these muscles went flat or limp and became permanently unusable. For all the muscles, the lightest-weight test was the last one performed.

Muscle lifting @ 220 mA
71.6 g
60.0 g
50.0 g
25.0 g
Muscle 1: Packed, no silicone
Muscle 2: Spread, no silicone
Muscle 3: Packed, silicone
Muscle 4: Spread, silicone

Thanks to the new test rig, I think these are more reliable than results I've posted previously – but you should still take them with a grain of salt, because running the tests spanned a hot summer afternoon, and I can't hold ambient temperature in the house constant. I wish I could repeat all of these many times and take an average, but I really don't have the time right now. So I'm putting up what results I have.

With that disclaimer out of the way – the plain old close-packed muscle without silicone is the best performer by a good margin. I wanted to see if the silicone coating would have any detrimental effects on the properties of the muscle, and it appears that it did … so even if it does help cut down on manufacturing failures, it's probably not a good choice. My awkward attempt to spread the coils on the annealing rod doesn't appear to have panned out well either. But that doesn't mean spread-coil muscles are entirely out of the question.

Perhaps one could wrap the muscle around the rod with close-packed coils – as it is naturally inclined to configure itself – and anneal it that way, achieving nice, even coils. Then the muscle could be removed from the rod, put under load and stretched a fixed distance, and annealed again by running an especially high current through the heating element. I suspect this would achieve much nicer results.

Lots of failure in this post, in that none of my little “improvements” really worked out for the better – but maybe someone else can avoid the same dead ends.

Until the next cycle,


Sunday, January 4, 2015

On Dragon Age and Hard Choices

Woo, long blog hiatus! I was busy moving to a new house, and then along came the holidays, so I haven't had much time for robotics or the blog lately. But somehow, I found time to play Dragon Age in the middle of all that. No, I don't mean DA: Inquisition, I mean DA: Origins. I am VERY late to the party. But I suppose with the new game out, this is as good a time as any to talk about where the series started. (This article has mild spoilers.  If you haven't played, I think you can read it without ruining anything.)

DA:O did an excellent job of creating a living fantasy world that I actually felt part of. I could believe myself a Grey Warden, a powerful figure sworn to protect the world from an ancient evil. But the part that has really stuck with me from this story is the amount of free will I had as an actor in it, its propensity for pushing me to make choices in nuanced situations, and its mature take on the results. I'm used to wish-fulfillment from this genre -- games that let me craft the perfect hero story, and if they include much tragedy at all, make it inevitable and not my fault. DA:O is a bit different.

I was kind of naive at this point.
Example: this world has golems in it. Awesome. As I dug deeper, I found that the way they are made in this setting is ... not awesome. Eventually, I was faced with a choice: preserve this dread technology and hope against past experience that someone would use it wisely, or discard it, possibly eliminating golems from the world forever. I opted to destroy the means of golem creation. As much as I love golems for reasons of both personal fancy and practicality, there were more important things to preserve here. And this was not the most painful decision the story required me to make, by far.

Thanks to a combination of numerous options and different ways for players to interpret the story, the endgame of DA:O is going to be unique for every person. For me, the path I had followed through earlier parts made it especially pivotal, and having put emotional hooks in me, the plot seemed intent on exploiting them. I was offered the perfect fairy-tale ending ... the one I had been hoping for ... at the price of a couple of sketchy choices. Alternately, I could do what I thought was truly best for the world I was trying to save, but only by giving up what I wanted most in that world. I chose the latter, and the consequences landed. There was no miraculous deliverance, no last-minute "power of true love" or "good karma" or surprise rescue to fix everything. It was miserable and incredible and different from any other game experience I've had. Dragon Age actually made me pay to be a hero, and the payment was my choice.

Whether literally or figuratively, real heroes bleed.  A lot.
Some part of me wants to replay the game as a slightly less scrupulous character. I could throw self-sacrifice to the winds and get that ending I really wanted. "It's just a video game, you know," says this part of me. "You can do that. You won't actually hurt anything." But whenever I start thinking this way, a second part of me warns that doing this would cheapen the whole experience. As soon as it becomes "just a video game" rather than a world that I embrace on its own terms, I will have lost what really makes it meaningful. As soon as I stop thinking of it as a place where I make morally relevant choices, I won't be a Grey Warden any more. I'll be nothing but a petty escapist, playing at being a hero without caring what's actually involved.

And at the end of the day ... that's not what I want. What I want is to wear my heraldic griffon shirt and feel that in some tiny way, I earned it.

I haven't forgotten about the artificial muscles and other robotics projects, I swear.  Thanks for your patience.