Robots are learning to see, reason and move. The harder problem is what happens when they make contact with the real world.

Aaron Prather’s recent writing on robotics and touch prompted VeinTrain to look at a problem clinical skills educators have spent decades watching from the other direction.

Robotics researchers are asking how a machine learns to grip, feel resistance, detect slip and adjust its movement. Clinical skills education asks how a person learns to use sight, touch, position and pressure together, then make a correction measured in millimetres, sometimes fractions of a millimetre.

The two questions are getting closer.

Watch on Vimeo: The Flatpack Vein - Innovation Showcase

What do skilled hands know?

“My father was a chef. I remember the speed and precision of his knife skills, but speed was not really the point. His hands knew the work. They knew the weight of the knife, the resistance of the food and the tiny adjustment needed before conscious thought had time to describe it.”

A pianist’s hands reveal the same thing in a different form. Two hands do different work at the same time, following rhythm, sequence, pressure and sound. When Sofia played after breaking her wrist, the skill had not disappeared. She had to adapt it around a changed physical constraint. That is a useful reminder that expertise is not simply the repetition of a memorised movement. It is continuous sensing and adjustment.

Clinical hands work in the same way. Every patient, vein, device, angle, position and tissue response is slightly different. A practitioner does not simply replay a standard movement. The skilled hand reads what is happening and makes small corrections.

“I have spent decades watching hands. Now, for the first time, technology may allow us not only to watch them, but to begin to measure what they are doing.”

Why fractions of a millimetre matter

In cannulation, a movement can be technically recognisable and still be too large. The relationship between the needle, cannula, skin and vein changes quickly. A fraction of a millimetre may separate a controlled advance from loss of position.

This is why a checklist can never tell the whole story of skilled performance. Checklists are valuable. They help us structure preparation, safety and sequence. But they do not fully capture how firmly a hand stabilises, how an angle changes, when resistance is felt, or how quickly the practitioner responds to visual and tactile feedback.

Nor does completing a course, by itself, establish clinical competence. Learning needs an appropriate route into supervised practice, workplace assessment and local sign-off. The challenge for training is to give people enough high-quality physical practice, with feedback that helps them understand what their hands are actually doing.

What simulation has taught us

Clinical simulation has changed enormously over several decades. High-fidelity mannequins that could breathe, move and respond, some with price tags approaching £100,000, opened important possibilities for team practice, emergencies and complex scenarios. They also brought practical limits: cost, access, specialist facilities and limited time with the equipment.

Virtual reality has added another layer. It can place the learner inside a situation, provide repeatable scenarios and make decisions visible. It is particularly useful where context, judgement, communication or sequence matter.

But the more visually sophisticated a simulation becomes, the easier it is to assume that fidelity on the screen is the same as fidelity at the fingertips. It is not. For many practical procedures, learners still need to handle a real device, meet physical resistance, coordinate both hands and repeat the movement often enough for it to become controlled.

This is not an argument against high-fidelity simulation or VR. It is an argument for being clear about the learning job each tool is doing.

Simulation has become extraordinarily good at adding information around the learner. The next question is what happens at the fingertips.

Teaching through a camera

VeinTrain began using webcams for skills teaching in 2012. At the time, teaching a practical clinical skill through a camera felt unusual. It also revealed something important: a camera could bring the educator’s eye much closer to the movement.

Later, Zoom made it possible to magnify hands, devices and practice models in ways that a learner at the back of a room might never see. Remote teaching did not remove the need for physical practice. It changed what could be observed and discussed.

That distinction matters. The learner still needs something physical in front of them. The camera becomes another pair of eyes. It can show position, sequence and movement. With the right design, it may also help make feedback more consistent and timely.

From camera to AI Vein

This is where VeinTrain’s work on the flatpack vein, camera-supported practice and the emerging AI Vein concept becomes interesting.

The aim is not to make the most expensive simulator, or to replace an educator with an algorithm. It is to create a practical architecture around real hands and real physical practice.

That architecture could connect:

• a physical practice model that is accessible enough to use repeatedly;
• a camera that can see the hand, device and practice area;
• movement measurement that can identify position, angle and change over time;
• real-time prompts that help the learner notice and correct a movement;
• physical feedback from the model itself;
• an educator who can interpret performance in context; and
• a governed route from learning into supervised workplace practice and assessment.

A wearable camera or glasses may eventually add the practitioner’s own viewpoint. That creates new possibilities, but also new responsibilities. Any serious system would need careful work on measurement definitions, validation, privacy, governance, safety and the limits of automated feedback. A promising prototype is not yet a validated assessment system.

Can technology measure adaptation, not just completion?

The most interesting question is not whether a learner completed the procedure. It is whether we can begin to see the adjustments that constitute skill.

Did the hand detect a change and respond? Was the correction controlled? Did the learner maintain stability while another part of the movement changed? Could feedback help them recognise a pattern that is difficult to describe in words?

Robotics is approaching the same frontier from the opposite direction. Vision alone is not enough. A robot operating in the physical world needs information about contact, force, slip and uncertainty, then it needs to act on that information safely. Aaron’s phrase, “the year of the hand”, captures the shift neatly. The difficult work is moving from impressive movement to reliable interaction.

Humans already have extraordinary tactile and adaptive capability. The challenge for clinical education is to understand how that capability is learned, how it can be observed and how more people can gain sufficient, purposeful practice.

It took decades to get to this simple

The flatpack vein looks simple. A camera can look ordinary. A short prompt may appear obvious. But simple, usable learning architecture is often the product of years of watching what learners struggle with, what educators need to see, and what expensive technology still does not solve.

The future of clinical simulation may not be a choice between virtual and physical. It may be an intelligent architecture connecting both.

Robots are learning to touch at the same time that AI is becoming capable of watching human hands. Somewhere between those developments is a new opportunity for clinical skills education: not technology for its own sake, but better access to physical practice, more precise feedback and a clearer view of how skilled hands learn.

Explore more in the Clinical & Learning Toolkits, or see how VeinTrain connects practical tools with learning through Learning with VeinTrain.

Further reading

This reflection was prompted by Aaron Prather’s writing on robotic dexterity and tactile sensing, including The Year of the Hand and Manipulation Is the New Frontier.

With thanks to Aaron Prather for opening this line of thought.