Angular constraints of prosthetic grippers and functional success correlation [technical evaluation]
I had evaluated, subjectively, the grip performance of various prosthetic options that I have. These have been already analysed in the context of grip taxonomy, where so far, research has largely focused on grip geometry as such, using some idiosyncratic logic that I found not too relevant.
Using a more relevant logic, I approached the question of grip mechanic from a different angle, both verbally and proverbially speaking: from a user angle, both actually geometrically and subjectively speaking.
I realized that most of my frequently used grips and grip situations fall into a far more narrow range of angle distributions than I had ever assumed. So I sat down to add "typical object angles" to my already present grip success statistics over a list of my most frequently or typically used grips. Then I did that in theory and THEN I figured, why not go and video some. Thereby, a prosthetic hook as gripper device appears to be a lot more advanced, design wise, geometrically, in reducing device materials, bulk and design to approximate a really good overall use performance than the iLimb (which I have here also for as much testing as I like) and with that, many current commercial (or other) multi articulated hands.
In fact, prosthetic hands appear to be by far the older (and thus possibly less reflected) geometric design idea of a prosthetic arm's terminal device than the definitely more modern split hook. I may also go history hunting, but the claim that a split hook is old or outdated, and that therefore by inference a prosthetic hand is automatically new or more modern, as an idea, is wrong, particularly technically speaking. But also historically, to replace a hand with a hand is a straightforward design idea, that does not take any imagination, thus it is reportedly old, very old.
The far more elegant reduction, also of angles and controls, to fit into the limited action and option constraints of an arm amputee, is certainly that of a body powered split hook. It boils down the prosthetic needs to a successful sleek elegant reduction of a functional minimum, making it the ideal choice for anyone that wants a maximum of performance from a minimum of failure, cost, decay, bulk, futile grip attempts and total overhead. The subtle distinction is that a "body powered split hook" is an entirely different beast than a passive hook, obviously, which probably no one ever noticed, particularly not the people that assumed that a body powered split hook is best portrayed by installing a "Captain Hook" metaphor.
- Being constrained and not free to move generally
- Being constrained and not free to move with a prosthetic arm as right below elbow amputee
- Grip constraints
- The post grip wiggle (PGW)
- Angular constraints of Hosmer hook 5
- iLimb Touchbionic
- To be added [work in progress]
- Typical object angles and grip quality
- Predicting constrained prosthetic arm grip quality from average or typical object angles
Being constrained and not free to move generally
A person often is nailed, constrained, fixed to certain positions. There are instances when a person is not free to move in a 360-degree spherical way. This struck me as very relevant a long time ago: in 1998, I spent time with a friend in the Swiss mountains. He tried to sell a golf course architecture to a town, where a local architect had already drawn up a different golf course plan. As that local architect appeared to be biased towards a competitive golf course nearby, my friend was not sure how much bad planning was part of that golf course part. The town people at that time seemd to lean more towards the local architect but my friend figured that he needed to avoid a planning catastrophe. Also they obviously wanted that contract. I quickly identifed scenic views in that area as the major aspect for tourist sales and for alpine happiness. I had just spent a year in Australia, playing golf as newbie, and I was struck by what golf does to you, particularly in terms of that. The better the view from the golf course the way it is built, the better the scenic experience. There are golf courses in Melbourne where the tee-off forces you to stare at a tall metallic container as part of an aluminum shmelter or oil raffinery or such - and if you are into playing "industrial game levels", this is absolute heaven. It resets your brain. The sun shines, the metallic structures reflect, and you are constrained to enjoy them. You cannot wiggle out of the visual grip there. It makes you stop and look again. And similarly to that experience, Swiss scenery has it, though, that is not homogenously scattered throughout a spherical 360-degree viewport. No! We have totally unsightly stuff to see - boring parts, ugly parts even. From a lot of army training in these Swiss mountains, I also knew that if you are constrained to stare into a rocky slope all afternoon, because for some reason you have to do some weapon shooting training, you will be constrained to stare at a rocky slope for hours and despite all that scenery, you will not clock many hours of actually seeing, looking at and enjoying true alpine scenery. A golf course is no different: it will constrain your view to a small field of view when teeing off, where you will aim your ball to the fairway with maybe some 10-15 degrees width but not more. You will spend time then approaching the green, with similar constraints - only looking in a set direction. The golf course was to be built close to the main valley road, and it is entirely possible to generate a set of views only consisting of road, piled up rocks, rocky slopes, boring forest parts and such, without ever seeing a sunset, far away snow covered mountain tops, the beautiful churches of the villages nearby, and so on. And with that in mind, I performed a side-by-side review of both proposed golf course architectures, which I used to draft a detailed report where both submissions were placed side by side with respect to the constrained player views, then rating the scenery - with the direct effect that my friend was successful and his golf course architect won that round. That golf course to this day obtains top ratings particularly also because the setup basically forces you to visually consume great landscape rather than boring in-betweens such as staring at forests or rocky parts - which, trust me, there are plenty as well. Only now, the golf course guest is kindly invited, constrained, force, to consume nice scenery.
So, with regard to an intellectual understanding how to intelligently build a constraining architecture, I may be not exactly new. I thought that up myself to begin with twenty years ago.
Being constrained and not free to move with a prosthetic arm as right below elbow amputee
The grip that I can perform with a prosthetic arm (that has a terminal device) is constrained. If one defines the grip as a plane, given by a plastic card or cardboard gripped in the gripper, the range of angles that this plane can take when I move my elbow, shoulder and even trunk is constrained.
Most objects in everyday life are constrained as to their typically appearing object part angles that are accessible to gripping. Thus, with relation to three planes, normal or typical object angles can be given as follows (axis: typical defining axis for grip, i.e., most longitudinal axis etc.; a= vertical plane, b = flat plane, c = wall plane; I used decimal degrees):
- bread loaf - a bread loaf is usually at table position and horizontally placed before the person cutting it, and in fact bread loaves step into appearance as ADL-defining item in order to be cut into slices; a bread loaf has the angles a=90, b=0, c=0.
- pen - to be used for writing, a pen is usually held in a constrained way (a=0-30, b=60, c=70).
- keyboard key - the typical keyboard key is almost entirely flat with relation to a horizontal absolute, so you could approximate this with a=90, b=0, c=90.
- typical handle bar of a striker or gurney: a=90, b=0, c=0.
The post grip wiggle (PGW)
The post grip wiggle (PGW) in an unnerving geometric position change of an object that is dragged from an inert or prior position into a gripped position, where it is held by a prosthetic arm. Due to angular constraints, the original position may not be preserved so the object experiences a jerk-like re-positioning that I will term PGW.
The PGW is a watershed phenomenon that seems to occur in between angular grip/object matches, and mismatches.
One aspect here is that for a hook, the following occurs:
- Object is gripped -> Hook aligns itself with object (snug fit) -> Object is lifted and no wiggle occurs
Whereas with prosthetic hands, due to their rounded finger tips, this is a typical observation:
- Object is gripped -> There is preliminary fit of gripper to object -> During / after lifting, the finger shape finish alignment, thus changing angles, in some instances the forces change the finger position -> Object is lifted in a final grip constellation
Angular constraints of Hosmer hook 5
The great use of a prosthetic hook, actually, of the Hosmer model 5, in every day life, for ADL so to say, lies in its mostly effortless angular fit for many differently oriented objects.
The cup rim at the left and right side align with the split hook's grip slit but not the top or bottom margin, of what geometrically appears to be a round shape. There is a bit of PGW on the second lift as there, the angle of the cup is not totally aligned with the hook as it happens to be angled right there. I can rotate the wrist if I want to, but as a matter of fact, most of the time, no one cares about tiny PGW when grabbing an empty cup.
Also the handle of the cup effortlessly aligns with the hook grip slit and thus the cup can be grasped with a perfect grip quality. There is a small post grip wiggle (PGW) that tilts the cup slightly to the right if you look carefully. If the cup is empty, who cares - but for a full cup this is a decisive aspect, the fuller the decisivier (engl.?).
Next image here, with just a bit of elbow bending, the hook grip split is aligned horizontally, thus effectively grasping the cup handle sideways. This also is part of a range of grips that lie within the boundaries of matched grip/object angles.Minimal PGW with a minimal upward tilt.
Next image, plates from dishwasher into cupboard ("cup"board? ah, never mind). In fact, the plate position in the dish washer just as the storage position of the plate in the cupboard both lie well within the angular range of what a Hosmer model 5 hook covers with great ease and perfect object/gripper angle matching. It is that type of ease of access that really goes very far when reflecting where all that added grip angst comes from that ends up defining the End Of The Day Feeling (EODF).
Here is me gripping the handle of an upside down pot, one of the many benchmark items of ADL that are not contained in SHAP oder Cybathlon repertoires.
This plastic container gripping is interesting: here, the slightly oblique container wall aligns with my forearem and hook grasp split so no major post grasp wiggle is observed.
The same is true for the top wall of the container.
Here, however, a small effect of post grasp wiggle is shown as the alignment of the container wall is slighty different from the prosthetic hook, so at the moment the fingers close in, the object swivels a few degrees anti-clockwise (if seen vertically from the top). These minute wiggles are a major aspect in prosthetic hand grasping. They are a big difference when comparing hooks and hands, from an everyday use perspective.
The versatile hook shape also easily lifts out this upside down container, another benchmark shape not contained in SHAP or Cybathlon repertoires, without significant post-grasp wiggle (PGW).
This shows lift-out of a plate with no angular grip problems such as PGW.
Also, a plastic bottle is retrieved without PGW.
Grasping a metal brush used for waxing skis is a simple reliable one-motion sweep task.
The grip geometry of a hook versus a simple plastic battery charger is dream like. A single sweep and the grip works.
The Hosmer 5 hook incurs only minimal PGW to the flat, horizontallz positioned metal scraper.
A simple metal brush used to wax skis is a seemingly extremely simple object to pick up. All the same, an iLimb as mounted on a normal socket does not have a grip angle that allows to easily pick up the brush.
Here, the thumb intends to grab the object at first but its further oppose and grip close angle bypasses the object. The iLimb hand thus is awkwardly designed with respect to the angles actual everyday objects have most of their time.
Slightly bending my knees lowers the thumb to better access the brush edges, but, still no luck.
The fingers do not manage to get a perpendicular form closure and the brush falls out despite best effort.
Compared to the attempt to approximate the difficult angles of the grip space an iLimb offers with a body powered hook, it is fair to say that with respect to grip efficiency, the hook is the geometrically far more advanced prosthetic terminal device.
Trying to pick up another rectangular seemingly simple object is daunting when the angular grip constraints of the iLimb mismatch the object's positional angular requirements by too much.
Here, thumb and index promise to meet each other upon start of grip but end up not doing so.
Here, the grip is inefficient in that it pushes the charger out of the hand while closing the fingers. Typical angular mismatch problem.
Another attempt that illustrates how just somewhat inconvenient the angular constraints of an iLimb is vice versa a world with typical, frequent object angles.
The reliable orthogonal form closure of the fingers towards the object to be gripped just seems to be missed by a bit, once again a failed grip attempt.
Finally, some type of grip works out.
Compared to using an iLimb, the grip geometry of a hook versus a simple plastic battery charger is dream like. A single sweep and the grip works.
A hook is able to just grip the cup without any "attempts" or added complications such as the various attempts with an iLimb bring about.
The iLimb incurs significant PGW (post-grip wiggle) to the metal scraper here: the original horizontal position of the scraper is clearly changed to some 40 degree angle after the grip is complete. The reason is to be sought in the rounded finger tips in absence of well performed finger motion coordination.
The Hosmer 5 hook incurs only minimal PGW to the flat, horizontallz positioned metal scraper.
Another attempt to lift the metal scraper with the hook: minimal PGW.
Due to a poor thumb stabilisation, the cup lift with the iLimb horribly derails.
This is what we want to get from an iLimb: top performance, grip wise. However, due to real grip constraints caused by objects naturally occurring at different angles, this is not encountered too often.
A hook is able to just grip the cup without any "attempts" or added complications such as the various attempts with an iLimb bring about.
Try to lift a simple ski wax package. The object is at odds with the angular constraints compared to the constraints that the iLimb has in terms of ideal grip angles.
Trying a different approach, the thumb motion appears to intend a correct grip but fails ultimately.
Here, the iLimb angles finally result in a partial grip so the object is lifted, not too reliably but still.
Geometry of precision grip of iLimb is off
Precision grip is not just out of angle, it is out of touch for small items.
Difficult gripper vs. object angle constraints for iLimb
The fabric requires significant bending to fit the grip angle of the iLimb thumb/index finger grip, which, on top of it, is off.
The angle between the clothespin and the hand is entirely off, so no grip can ensue.
Instability of within-grip geometries of iLimb
Precision grip malalignment of iLimb
With the geometries of the iLimb grip angle being peculiar or different to begin with, issues are further complicated by a precision grip that even without object touch derails, unaligns itself.
It is fair to say that the iLimb, out of the box, will be impossible to pilot reliably.
That is not saying it cannot be done. But: professional, really professional, mad pilot skillz are required to pilot this device across any parcours including daily life. Having it on the arm for a few days is not enough. It needs to be worked in, applied.
To be added [work in progress]
- Hosmer hook 6
- Becker hand
- TRS Prehensor
- TRS Equilux
- Toughware Equilux
Typical object angles and grip quality
Typical objects are at a given angle, or within a range of certain angles, in a rather constrained range.These angles appear to at least partially determine the grip quality, whereas that is dependent on the actual gripper or terminal device.
A grip is, depending on the angle constraints, one of these three:
- perfect and perfectly stable, with only minimal PGW
- nudged, approximated, often with significant PGW
- difficult or actually impossible
Predicting constrained prosthetic arm grip quality from average or typical object angles
I defined object angles as angle of most prominent longitudinal axis to a vertical, flat and wall plane. Then I typed in the most frequently encountered angles for a total of ~41 single grips that are of various relevance throughout my days. These were added in almost a blinded way with regard to grip quality of various devices: already over a year ago, I had classified the grip qualities for my terminal devices along a 1-10 scale, worst (1) to best (10), and not considered these when evaluating grip angles.
The average object angles (a,b and a,c; with three angles a,b,c, that makes two diagrams) thus can be used to mathematically "predict" the grip quality for a given prosthetic option, and through use of the mathematical model, a predictor shape results. That shape, a 3D-surface, will be higher (with spatial diagram, right side of below image) depending on average object angle where grip quality is better. I used a Neural Net model (as that apparently has no problems with multivariate collinearity or non-normal distribution). As the curves are iso-values across the 3D-surfaces, the dotted side points to where the surface's slope goes up (i.e., to which direction the grip quality gets better).
As the data that I used is a bit overly specific, I evaluated it as such (top images in the image here), but I also evaluated a massively bootstrapped data (n=625 000, added noise 1 standard deviation). The differences are not extreme between the raw and bootstrapped noise-added data.
And while that is not even too interesting, the contour curves that characterize these surface shapes are relevant as they allow to compare various grip modes or terminal devices against each other. Content-wise, the result of this clear: the average object angles are ideal for some, but not for other, grips, depending on the specific terminal device that is used. I also used the setup to predict the non-failure-requirement contained in the same statistical cohort of constrained typical object angles (see diagram curves at the bottom).
The geometry of the prosthetic gripper plays a considerable role in determining grip qualities across (at least my own) ADL / work grasps, simply because most typical object angles are constrained. Various terminal devices, or not wearing one, will define a preferred or ideal range of action, that also contains constrained access angles. The conflict between constrained access angle and typical object angle results in anything between a perfect and absent grip quality. This is clearly illustrated in various video sequences (see above).
The grips where failure is not an option can also be characterized through their object angle constraints. Then, ideal function across angle space can be compared between grippers, and checked against requirements for no-fail-relevance - if you read the diagram, above, that is.
Wearing a prosthetic arm may be optional if one performs grips where failure is an option or where grip failure only causes no or negligible cost. Where grip failure is absolutely intolerable, there, a prosthetic arm's performance with respect to these is critical however. A prosthesis that fails grips who require a reliable function is, mildly put, toast. A prosthesis that fails grips that are not necessary is not toast but still useless. So to understand, addressing and succeeding in the domain of constrained grip angles must be absolute key for anyone that wishes to provide grip performance to arm amputees. Introducing variable wrists is a solution but they can be notoriously cumbersome, if not also heavy - so think very, very hard before considering to go gah gah there. We built a perfect wrist for right below elbow amputee function ourselves, and that was not easy.
Turns out that the classic hook devices almost perfectly align with the geometry of relevant, i.e., non-failure-relevant, grips. The Toughware Equilux seems to take a middle position, geometrically, between TRS Prehensor and voluntary opening conventional hook devices which is logical based on its hybrid control system. Not wearing a prosthesis at all appears to cover a part of the non-fail-relevant grip angles as well based on geometry alone, whereas the constrained angles of an iLimb seem to be mostly at odds with the angles found in typical objects that require reliable grips. This is easily illustrated, see above for imagery. In fact, wearing no prosthesis covers more angles of the few relevant grips that it performs well than wearing an iLimb, which confirms anecdotal experiences, while the actual rating of grip quality as an overall rating from my view.
If anyone wants to build a multicarticulated prosthetic hand that conforms to everyday life requirements, one could start at considering the bare realities that are easy to assess. A very repetitive grip action with a pronated hand may result in lateral epicondylitis, which is why for some evolutionary anthropology reasons, our cultural object space seems to favor a non-pronated and non-supinated precision or power grip a more frequent across many ADL and work situations - and if it does not, you should look to it that it does. The old pincer gripping Otto Bock and Hufner hands offered a more useful grip with regard to angular constraints than the modern multiarticulated hands seem to offer.
As far as I see it, this consideration has not been performed in any prosthetic hand design so far. I mean, who really crawls deep into angle space to really consider why a prosthetic hook really feels so much better after stressful days at work? Certainly no one in R&D. It is us, the users, that harbor that type of knowledge that is increasingly ignored by R&D. The fact that anthropomorphic prosthetic hands are not accepted too well, or, even if they are bought and purchased, not used too often, may be found in their insufficient grip geometry alignment with everyday objects. Split hooks have evolved around the everyday use space for a considerable time now and have to be regarded as "that shape that grasps most of the objects most comfortably".
Look at it the other way around: isn't the prized question your were too shy to ask that hooks are so "ugly", why are they still around? Solve that puzzle yourself - or re-read this, which may be one way of solving it.
"Bionic" prostheses Activities of Daily Living and Sports Algorithm regarding best choice of prosthetic arm Artwork and Do It yourself Corner Asymmetry Bimanual task benchmark Cable control Design principles Grip testing Health Hosmer Dorrance Investors' Corner Mechanical works Myoelectric prostheses Postural interference myoelectric control problem Prosthetic hand Prosthetic hook Prosthetics Research tips Specifications and comparisons Touch Bionics iLimb