Playing Yenga at a prosthetic arm workshop was interesting.
The hightech iLimb hand (Touchbionics) as all "bionic" prostheses are presented as clever, intelligent or sophisticated. Really, the hands are highly sophisticated, weak but funny, and should be "controlled" with just two signals (one for "open", one for "close") and that invariably causes issues. The increasing "control" options built into these hands now amount to an unrecoverable mess of finger combinations termed "grip patterns", who can neither work (link) nor be switched reliably by muscle signals - as one example, they switch grip while carrying a trash bag or hanging clothes (link). Really one has to become one big focused non-distracted puppet master that never sweats in order to at least approximate good control.
Company representatives that spend a lot of their work time for training are just somewhat good, and suffer from chafed stumps just as anyone - without even having actually lifted hard or worked their butts off physically; so the #voightkampff dilemma remains (link). When asked how to change a bike tire, the i-Limb ambassador said to take the tire to the bike shop. Of course I can do that without any prosthetic arm. And it is not a surprise - i-Limb and other "bionic" hands are first and foremost symbols, they are representing tools to show case membership of the leisure class (link) not working items that are truly meant to be worked (but instead designed edgily (!) to tear down gloves within minutes) - and so, actually "using" them desecrates that actual societal symbolic meaning just as using a clean fresh smelling paper handkerchief is never to be used as "snot rag", and just as a fast speed-promising sports car such as a Porsche is never to be actually used to drive aggressively (as then it loses its symbolic privilege). So to assume that an i-Limb is meant to be actually used for grip purposes is a misunderstanding. Much rather, it is a hand looking device meant to exhibit gadget properties as symbolic sign of coolness that falls apart already when looking at it. It contains all aspects of a luxury item. As Touchbionics now offer "model upgrades" for their hands, but at a probably very high price, without having clarified their glove situation disaster (link), I felt like wanting to first dissect and analyze the relevant grip mechanics, and the Yenga game is just great for that. After all, my existing i-Limb already suffices as symbol of representation, as device to show around; the current problems to be solved are grip reliability, glove durability and electrode / software reliability; no obvious need to further add to the damage.
So, taking a step back. Analysis time, grips.
- Baseline - Yenga tower played with Hosmer Model 5XA
- Baseline - Yenga tower played with arm stump and a screw driver
- Test - Yenga tower played by Becker Imperial hand and Centri PVC glove
- Test - Yenga tower played with Hosmer Mod. 555X hook
- Test - Yenga tower played with Hosmer Mod. 6 hook
- Test - Yenga tower played with iLimb Touchbionics
- Why/how do prosthetic terminal device grips or alternative approaches tilt the Yenga block / object?
- Summary of Yenga exercise
- Hook versus "bionic" - Hosmer 5XA, Hosmer 555X and iLimb
- Hook comparison - Hosmer 5XA, 555X, 6
- Stump, Becker hand or iLimb
- All against all - hosmer 555X, Becker hand, iLimb or bare stump (without tool)
- Research for anthropomorphic hands
- Real life Yenga - cleaning out old harddisk collections
- And after all has been said and done
Baseline - Yenga tower played with Hosmer Model 5XA
Observations: the performance for precision pushing is superb, obviously. The push function also rules when using that hook for typing on a computer keyboard. At times however, tilted angles of the prosthetic arm coincide with the slanted hook shape, causing the final phase of the grip closing in of the nitrile covered (deformable) inner hook surfaces with the wooden block to result in a minimal tilt.
However, Yenga as a game is destined to showcase subtle grip problems, and further below in this post I dive into exactly why the Hosmer model 5XA has this "final phase grip tilt" as a feature (not as a bug).
A really good moment in the video is @1:49 where exactly that can be observed: when settling in on a firm grip, the wooden block suffers a minimal tilt which makes the tower shake a little.
Control strategy remarks: separating postural and approach control, opening / closing control and maximal visual accessibility is easy with this body powered arm.
Baseline - Yenga tower played with arm stump and a screw driver
Observations: the screw driver, particularly this one featuring a rubberized metal rod, is a great tool to extend my hosed arm for the Yenga game.
Its control is leveraged by gripping it with my elbow.
With very soft deformable skin surfaces, an elbow inside has an extreme grip performance relative to its low shearing forces and possibly power to hold smaller objects (for larger objects such as exceeding 3-4 cm in size, the elbow is extremely powerful).
Watching the video closely, one may quite possibly understand for the first time exactly why "not wearing a prosthetic arm at all" does not at all cause my manual and even bimanual performance to be all that significantly lower.
If anything you will want to most seriously revise your understanding of grip mechanics and how grip problems are *actually* solved (link).
Control strategy remarks: postural control and grip are very intuitive; slip control is near perfect as I sense and see everything very well. If you allow for different approaches to grip, this should be the option you consider first.
Test - Yenga tower played by Becker Imperial hand and Centri PVC glove
Observations: despite its apparently clunky shape, the Becker hand has a locked thumb and index finger tip precision grip that is unparalleled in the world of prosthetic hands in terms of practical use.
The grip is repeatable in that each and every time the hand opens and closes, the index / thumb opposing geometry is stable and invariable. In particular, that precision grip contains no shear, as thumb and index move within one plane.
That also makes oblique applications of that grip foreseeable and thus, the subtlety of grip-object interaction becomes highly controllable in a very intuitive way.
The grip at @0:41 shows very well how precise and straight forward using this hand is.
Postural issues can lead to a wiggle issue as shown after @0:07, but these are not grip issues. The covering PVC glove also offers great grip object interaction with deformation occuring during the final grip phase.
Control strategy remarks: the same control strategy as all body powered devices applies here.
Test - Yenga tower played with Hosmer Mod. 555X hook
Observations: as opposed to the other hooks and the iLimb, the Hosmer Mod. 555X hook moves parallel and symmetrically shaped hook grip surfaces that are also nitrile covered (and thus deformable) so that the grip-object interaction here is extremely well plannable.
Final grip phase or lock-in wiggle issues with the Yenga game are not a tangible problem with this hook.
If you watch this and then still have to ask, why people still wear hooks, urgently mail me and tell me what prosthetic hand has a better grip than this. I really want to know. Thanks ahead!
Test - Yenga tower played with Hosmer Mod. 6 hook
Observations: I intentionally equipped this back locking slanted ultra heavy work tool Hosmer model 6 with added silicone tubes to make the hook surfaces seriously deformable on top of being slanted. Technically, this should result in a really wiggly final grip-object close-in phase with a lot of wiggle. Of course, that type of hook is geared towards holding and operating relatively heavy machines over a longer period of time, such as garden tools or so - but here, we are studying grip mechanics. The not so easy to understand but (below) logical issue becomes visible when watching @0:40 onwards where one can see that this grip seriously tilts the wooden block so it almost jams.
Control strategy: despite a perfect posture / grip separation for controls (as shown in the other body powered demos here) this device performs quite bad. This also illustrates that control paradigm and terminal device should be separated when evaluating devices.
Test - Yenga tower played with iLimb Touchbionics
Observations: after @0:43, the postural interference with flexed elbow and hand controls becomes a real issue; the fact that the i-Limb technically does not at all control thumb / index finger relative positions becomes a real issue after @01:02 when one sees how the "precision" grip forces a high degree of object tilt that is unavoidable. The real reason technically is that the i-Limb forces the thumb and index finger motor to run simultaneously while the fingers are at unplanned relative positions; opening and closing the hand partially will further deteriorate the "precision grip" geometry, overall resulting in "unplanned slanted" grip situations which is also reflected for everyday situations.
There are certainly some ways to plan a grip, but they are similar to taking the freeway around town to get to a shop that is 200m away. They require grip switches, resetting the hand grip position by fully opening it, and so on. The exact hand geometry then still is off for reasons of partial open/ close issues (link).
Control strategy remarks: posture and grip control interfere with myoelectric arms which is a known and unresolved issue (link).
Why/how do prosthetic terminal device grips or alternative approaches tilt the Yenga block / object?
It is worthy of investigation why and how that Yenga tower issue (wiggling / shaking) occurs in context of slight object tilt upon the final grip phase happening.
Why/how does the 5XA tilt objects?
Stop motion image of slanted hook gripping small object off a surface (below): you can see that there is a minimal object roll during the last grip phase with this slanted hook. The feature here is that this allows the cable opened hook claw to exert a "loading" force component on the object so it can be better picked from a flat surface. The effect is not pronounced but not more is needed. Top much slanted action results in less control over the object, as the Yenga use shows.
Why/how does the 555X *not* tilt objects?
The geometric design of the Hosmer model 555X is geared specifically for high precision grips. The symmetric hook design virtually avoids any object tilt when aligned properly.
Why/how does the Becker hand *not* tilt objects?
The Becker hand manages to avoid shearing motion between thumb and index finger tip - despite larger visual occlusion. That is probably the reason why it works so damned well.
Why/how does the Hosmer Mod. 6 tilt objects?
The Hosmer mod. 6 is definitely not a precision work tool. It covers the clean up, hammer, nail, mount and heavy machinery type range of applications. For that, a heavy slant is helpful in that it effects fast pickup of objects off flat surfaces.
The video between @0:06 and @0:07 shows how the object tilts at the very moment the hook surfaces close in on it and when the silicone tubing on the hook claws start to deform around the object to be gripped.
How does the iLimb tilt objects?
The iLimb was not designed or built as a grip precision tool. In more than one axis, grip surface shearing remains outside of actual control - no index thumb alignment, no tip contact reliability, and no way of telling which finger tip hits the object first and which tip surface travels into what *exact* direction. This amounts to the situation of only crude grip planning, which is acceptable once one knows this.
At any rate, if one were to design a precision gripper, and if it had to be a prosthetic hand with electric motors, taking some aspects for actual grip mechanics onto the design bench could be helpful.
Summary of Yenga exercise
A cost - performance - versatility assessment now will show the side by side aspects of detailed grip aspects between various devices.
Hook versus "bionic" - Hosmer 5XA, Hosmer 555X and iLimb
The workshop organizer for the iLimb that I recently visited specifically forbid me to unpack and use my hook in the process of the Yenga demonstration there. As the workshop organizer did this as a company deal (and not as prosthetic technician), they were biased.
So it is clear that there are deeper issues. When one parameterizes the actual Yenga performance, this is sky clear. Hooks
For small object manipulation, both hooks are massively better than the iLimb. They represent a true threat.
Hook comparison - Hosmer 5XA, 555X, 6
Among the top performing devices for Yenga I played with three hooks with different setups. And yet, people ask what hook to buy for everyday use.
The Hosmer 6 is a dedicated specialized high load larger object focused work hook not so useful for more intricated precision work. It is a great add-on but not too helpful for everyday use.
The Hosmer 5XA is the most versatile in that it also has a somewhat slanted grip that helps to lift smaller things off a flat surface and that helps holding a drink at a slight tilt without having to rotate the wrist - but other than that, the Hosmer 555X is the best buy for lightweight everday precision.
Stump, Becker hand or iLimb
Functionally for the Yenga game, wearing no prosthetic arm but squeezing a screwdriver into the elbow winds, hand(s) down. By far.
The Becker hand is a rather good approximation; if it has to look like a hand, let it be a Becker hand. The iLimb is great gadgetry but was probably not meant for this.
All against all - hosmer 555X, Becker hand, iLimb or bare stump (without tool)
Using the bare stump (without tool such as a screw driver), the situation is more realistic and complex.
The difficulty to grip an item is equally difficult to plan with the elbow or an i-Limb; wearing a Hosmer 555X will be a massive improvement there. Rigidity within a grip is worst without wearing a prosthesis at all but that weighs against a top performance in terms of comfort, absent added weight, affordability, the technical performance to do small corrections with the bare stump, and overall Yenga success even then.
If it has to have a hand shape, the Becker hand outperforms, but hooks are the way to go if one cares about grip issues.
Research for anthropomorphic hands
Unilateral transradial arm amputees can be really good at a lot of stuff without prosthetics on. Even with Yenga.
Two issues remain then:
- a really good precision grip, and, controls that achieve a maximally precise relative positioning of fingers towards avoiding shearing movements
- a really reliable "grab stuff for moving it around" grip, and a control system that does not just let go when under way
The Becker hand - if it must be an anthropomorphic device - solves both extremely well.
Without *any* surprise, prosthetic testing today neither tests amputee performance as is, or, using Becker hand or hooks - however, we can show that both are extremely good. Why that is becomes clear once one actually understand how to develop good prosthetic arms, and hooks as well as the Becker hand have seen, by now, at least 7 decades of very close iterations at relatively fast paces whereas most of the myoelectric industry representatives were probably far too arrogant to talk to demanding amputees. We know for a fact that companies reject requests to answer to amputees directly, and we can directly show how that behavior (link) directly connects to design failures (link) systematically.
- keep thumb and index fingers rigid, it is acceptable to bend the other fingers' joints though
- develop a useful surface shape and geometry for the precision grip
- avoid involuntary hand opening during transport of objects
Real life Yenga - cleaning out old harddisk collections
Fast grips required, wide grips, varying shapes, no fault tolerance for falls, very small objects (tiny screws) in at times hard to reach spaces, handling of small cables and plugs with varying resistances and thus requirement for a precise and fast, strong precision grip.
And after all has been said and done
After all has been said and done as it appears, iLimb Touchbionics marketing obviously saw the need to post this take on "playing Yenga" - apparently made by "Aktion Mensch" - across social media. The protagonist wearing the prosthetic arm is Bertolt Meyer who has been at it since many years and who also function as "Touch Bionics" ambassador.
No real life user that I know of can do that. At the workshop, we all struggled extremely. The controls blocked, the grip failed, and so on. Why? No one actually wears this for almost a decade to start being able to deliver that type of command. And why is that? Because even then - even running the iLimb with just its common adaptive grip - is problematic. iLimb is not only notoriously hard to control, it also contains a myriad of intricate problems that interconnect. It has intrinsic control issues that mess up the grasp, the balance is off, it messes up the costly gloves, it is loud, there is no strong grip and so on and so forth. Basically I do not know any job that easily allows *and* requires to wear an iLimb every day, and at least here, prosthetic arms mainly are bought by insurance to enable the user to work.
People do not understand though. I am not generally against good products. I am definitely in favor of good products! And so I would not post this if the iLimb - as a product - was 100% pure trash. In my view, it is so lovable, so cuddly, that one hopes that at least one day far away in the future, to have this hand "be" like a hand. Until then, though, troubleshooting keeps being worth the while.
If you are interested in more work related and difficult work applications, comparing body powered and myoelectric arm technology, here is an informative summary [link]
[corrected / revised article, June 30th 2016]