As my father, Gerhard Schweitzer, died in July 2026 at the age of 89 years [link], I should dedicate a page about our discussions regarding what we thought was a proper engineering approach, also as far as it relates to prosthetic arm/gripper design and functionality. It was not that we never talked about such or about related subjects, but it may very well be that this was not done in a similar way with other people / outsiders. Not all of his insights can be published, but some can. So here we go.

Rule 1 – do not perform prosthetic arm research

More than once, I had asked my father, why some, but not all, robotics institutes, have a prosthetic hand – as emblem, as symbol, as light house, as project, of futurisitic roboticism.

In a nutshell, his answer was that in his view, good robotics departments have far better projects than such.

He only hinted at the problem of too many constraints in the domain of prosthetic hands.

But he admitted that having such a device up for display for visitors and funding agencies had its own emotional aspects.

Not that he ever felt he should ever try that, too.

A good project does not have constraints that are impossible to overcome. A good project has a client side with stable and robust expectations. Without addressing that too explicitly, we agreed that neither is a criterium that prosthetic arm technology for amputees really fulfilled.

What later emerged, obviously, which I never discussed with him in any detail any more, was, that a body powered prosthetic arm with a direct very fast [link] steel cable control much like bicycles have [linkl clearly surpasses the myoelectric design catastrophe in so many ways.

That seemed rather clear from the outset.

He had grown up with an old greasy steel bicycle, and was working part time as car mechanic, at a time when he met and dated my mother. His occasionally dark greasy fingers at that time seemed to lend itself to an occasional remark, a laugh about old times, which was brought up every now and then. He got rid of his old bicycle only after our family had moved to Switzerland where the hilly landscape around Zurich and his time increasingly spent at work made that bike obsolete to him. At any rate, he definitely knew the immediacy and relevance of functioning brake cables.

Of course, cross-labeling army and industrial gripper research as prosthetic arm research does not fall under such – but other – precautions.

Of course, also, when I build my stuff that I wear: that is not some academic toodeling around trying out research ideas whilst trying to responsibly train future engineers. No: I run for my life doing that. My focus is build and (immediate) use, construct and survive. For me it is existential. Life here is no pony licking.

Rule 2 – whatever you do, use a good plan, and try to define a good model before doing practical steps (or after a few of these, just don’t wait for too long)

If ever, one needs a good idea of what one plans to do or what one is in the process of doing. This “good idea” thus could entail reflection, analytical thinking, as well as evidence based model building.

He instanced that many years ago, with his approach to a micro-processor controlled knee.

There, the idea of a Canadian research group to have such a knee joint provide better gait symmetry at the time apparently lacked a commercial product plan/basis, so Otto Bock financed this study where Daniel Zlatnik as doctoral student provided a well defined model and setup for such a knee joint. This provided the basis for the C-leg that, as we assume, was a commercial success. Then he published a piece on his website (now, www.mcgs.ch was deactivated), concerning a knee prosthesis [link]. This work also has since received numerous references in the scientific literature. The components, sensors and embedded computers used in modern prostheses have improved enormously since then, but the fundamental problem remains largely unresolved. The crucial issue is motor control: how an artificial limb can be controlled as naturally as a biological one. In the human body, movement depends on a continuous two-way exchange of information. The brain sends commands through peripheral nerves to the muscles, while sensory information travels in the opposite direction, informing the brain about position, force, movement and contact with the environment. In artificial limbs, this biological communication pathway still cannot simply be tapped into and used. Neural signals are difficult to access reliably, and their complex coding is still only partly understood. Together with Marie-Claude Hepp from brain research at the University of Zurich, he supervised a doctoral project investigating the neural control of hand muscles. The results apparently were frustrating: the problem proved considerably more difficult than the apparent ease with which a healthy person controls a hand would suggest. In the meantime, several Italian research groups have made further progress in understanding and interfacing with peripheral nerves. Nevertheless, a genuinely natural, robust and bidirectional connection between the human nervous system and an artificial limb essentially remains an unresolved situation. On top, amputees are mostly not in favor of invasive surgery with regard to such innovative risky procedures.

That does not mean one cannot approach a use case with a great model.

In many discussions, having a well laid out plan was a thing he talked about, pointed out, and that we often worked on also together.

At one point in time, some time in 1999, he decided that having a navigation system in his car was the thing to have. So he bought a VDO Dayton system, and had them put that into his Mercedes Benz 230. While that was not a fast car it sure was solid. During a road trip to Germany, I sat in the passenger seat playing with the ideal placement of the GPS antenna. Only once that location was found empirically, on the dashboard, under the windshield glass, did we reflect on possible reasons. The car specialist had said that GPS antenna was best placed in the glove compartment. However, they may have underestimated thick car body metal. Later I used that insight to better analyse my own GPS car function, then also in the light of solar storms [link].

At another point in time, he brought home the funny hand held gyroscope [link]. In our family I remember that we ended up marveling over that device’s max speed and forces exerted, and I even provided my father with an India ink drawing of the hand holding the gyroscope in Gerhard Schweitzer (1982) Antrieb eines Spielkreisels durch Taumelbewegungen seines Gehäuses, in: Festschrift zum 70.Geburtstag von Herrn Prof.Dr.rer.nat.K.Magnus [link]. To get there, he took the thing apart, and had someone at this institute replace the elegant plastic shell with a bulky metal housing, after which the device was almost impossible to use comfortably. Good research may require some sacrifice.

Not everything works out, obviously; some attempts fail to conform to its intended concept. But one then has to acknowledge that and draw logical consequences.

Ultimately, what I take away from many thoughtful reflections is that one wants to be rather systematic when approaching a new design, but also about one’s own serendipities.

Outlook

From his work emerged a much more speculative idea. What might a human being look like fifty years from now?

My father envisaged a future in which a human being might consist of more than 50 percent technical components — what he called technological transplants. Initially, such devices would be intended mainly to maintain or restore human functions: spectacles, hearing aids, artificial knees and hips, deep-brain stimulators for Parkinson’s disease, cardiac pacemakers and artificial limbs. But he saw no reason why technological development should stop at merely restoring normal function.

The next stage would be enhancement. Technology could provide humans with abilities that biology never gave them: infrared vision, artificial hearts capable of outperforming natural hearts, exoskeletons for extremely heavy physical work, or visual systems that automatically recognise faces and interpret facial expressions and emotional states. Which is funny given how humans, some humans, manage that well.

At this point, the distinction between therapy and enhancement would become increasingly blurred. Replacing a failing knee is readily accepted. Replacing a blind eye with an artificial one would likewise be understood as treatment. But what if the artificial eye could see infrared radiation or provide better vision than a healthy biological eye? What if an artificial limb became stronger, more precise and more durable than the limb it replaced?

For him, this raised a deeper question. At some stage, medicine might cease merely to repair the human body and begin deliberately redesigning it.

He saw an interesting parallel with the ethical debate surrounding xenotransplantation. Researchers were investigating whether animals, particularly pigs, could be genetically modified so that their organs might serve as replacements for human organs. Such research immediately raised uncomfortable questions about manipulating living organisms, crossing biological boundaries and producing animals specifically as sources of human replacement parts.

Technological transplants attracted less ethical attention.

A mechanical knee, pacemaker or neural implant are  less disturbing than an animal engineered to carry human-compatible organs from a range of considerations. Yet he argued that the ultimate direction could be remarkably similar. Whether biological structures were replaced by engineered animal organs or by increasingly sophisticated machines, the traditional boundary of the human body would gradually be altered.

Seen from this perspective, his prediction that humans might one day consist of more than 50 percent technological components was not simply science fiction. The process had already begun; the open question was how far it would go or if anyone would notice.

The cultural unease surrounding such a development was also nothing new. The homunculus, the Frankenstein myth and, much later, Hollywood’s Terminator all reflected versions of the same fascination and fear: the possibility that humans might construct, reconstruct and eventually technologically surpass themselves.

Fifty years into the future, the decisive issue might no longer be whether technology could replace parts of the human body. It might instead be whether there remained any meaningful boundary between repairing a human being, enhancing one, and ultimately constructing something that was no longer entirely biological — but was still considered human.

Conclusions

From a somewhat cool relaxed engineering view, a few things are a given.

General remarks

• Testing can be cautious at first but must be realistic and then hard.
• Not all older designs are worse [link]. Not all developments of “new” designs get better across decades [link].
• That is because disability does invite all sorts of unproductive and useless approaches, that are not appropriately double checked [e.g. link]. Researchers may even lack anatomic basic understanding [link].

Geometric constraints

• The prosthetic arm setup comes with unique geometric constraints all of which unfold, become manifest, upon movement or upon dynamic testing or use.
• Geometric landmark / registration point tracking using video capture from at least two angles is to be regarded as minimal documentation for any such project before any analysis is even considered. Unless you freestyle it as I do, but I am a first person investigator and user at the same time, none of all of y’all are will likely be in that situation. This is to be understood even before a first die is cast. This certainly includes the Cybathlon [link], howevermuch people advise there without any skin in the game [link]. This is relevant as having the wrong people on board will mess up the whole approach.
• The gripper itself imposes constraints [link]. Both macroscopic grip angle and concise gripper surface angles are relevant aspects.
• The arm and its controls also pose relevant constraints. There, body powered prostheses have to be built right before that is assessed, which is typically not done [link]. Myoelectric prostheses offer their own geometric use space constraints [link] [link].

Control constraints: reliability, errors/failure, speed

• Understand major and minor error causes for control errors, and do not mix them up.
• Understand intractable and fixable error causes for control errors, and do not ever mix these up  [link].
• Train your users accordingly.

Further aspects to approach systematically

• Material constraints
• Sustainability
• Comfort