Intelligent prostheses: from adaptive control to AI

Intelligent prostheses combine biomechanics, robotics and adaptive control and software to return maximum mobility and independence back to amputees. What used to be a purely mechanical replacement is now a sensor-supported, adaptive assistance system that recognizes movement intentions and translates them into almost natural movement sequences.

This shifts the focus from the mere fulfillment of functions by the prosthesis to active, safe and intuitive prosthetic support in professional and social life. What technical principles make this development possible, what does "bionic" mean in prosthetic technology – and when and how could AI be used? Our article shows some examples of current products, research and important manufacturers.

Arm prostheses are usually controlled myoelectrically: implanted or surface EMG electrodes detect muscle activity in the arm stump and translate these signals into movement commands. Advanced systems use pattern recognition to differentiate between complex muscle activities. One example is the Myo-Plus control system (including app) developed by Ottobock, which recognizes movement intentions based on trained signal profiles.

Such multi-articulating prosthetic systems enable differentiated gripping patterns, which makes everyday activities such as writing, cooking or operating technical devices easier.

In leg prostheses, integrated sensors register the knee angle, acceleration or load. Microprocessor-based knee joints such as the Genium system adapt damping and swing phase to gait speed and terrain in real time. Prosthetic systems such as the RHEO Knee from Össur use comparable adaptive control strategies.

This enables better mobility, for example when climbing stairs, walking backwards or in uneven terrain. Clinically, this results in improved gait parameters, higher walking speed and increased participation in the professional and social environment.

The term “bionic” describes the transfer of biological functional principles to technical systems. In prosthetic systems, for example, this refers to multi-articulating hand prostheses with individually motorized fingers, sensor-supported grip adjustment and programmable movement patterns. The aim is to achieve a functional approximation of physiological movement sequences through the prosthetic limb.

One example is the i-Limb Quantum prosthesis from Össur, which allows different grip types to be selected with software support. The bebionic hand, originally from RSLSteeper and now part of the Ottobock portfolio, also follows this concept.

With adaptive control, sensors measure and analyze biological and mechanical signals in the remaining limb in real time and control the prosthesis in such a way that it approximates the user's movement intention.

In the case of arm prostheses, this primarily involves the pattern recognition of myoelectric EMG signals in the limb, the classification of different muscle activities and the continuous learning of individual movement profiles to execute grip types precisely and reproducibly.

For leg prostheses, the software analyzes joint angles, acceleration and load. It recognizes gait phases and dynamically adapts damping and resistance parameters to different walking speeds or surfaces.

These functional principles are already evocative of the use of AI. However, genuine AI-based prostheses are currently still the subject of research, such as in Prof. Cristina Piazza's IN-HAND Lab at the Technical University of Munich.

Internationally relevant manufacturers of prosthetic limbs include:

• Ottobock (Germany) develops myoelectric arm prostheses with pattern recognition as well as microprocessor-controlled knee joints.
• Össur (Iceland) offers multi-articulating hand prostheses and adaptive knee joints such as the RHEO Knee.
• Open Bionics (UK) produces myoelectric, 3D-printed arm prostheses with software-based grip control.
• Vincent Systems (Germany) specializes in motorized hand prostheses with individually controllable finger modules.