Progettazione e realizzazione di protesi di piede del tipo ESR a rigidezza variabile

This dissertation presents the design, development, and validation of a new generation of energy-storing-and-releasing (ESR) prosthetic feet aimed at improving adaptability and personalization for active lower-limb amputees (K3/K4). Addressing the limitations of conventional fixed-stiffness ESR feet, this research advances a structured progression of passive, manually adjustable, and autonomous prosthetic architectures across multiple technology readiness levels. The first research objective investigates a passive multi-axial ESR foot, MyFlex-η, featuring an interchangeable elastomeric ankle joint capable of modulating stiffness in the sagittal, frontal, and transverse planes. Unlike conventional ESR feet that primarily allow sagittal motion, this architecture enables multi-planar compliance to better accommodate uneven terrain. Finite element modeling and bench-level validation demonstrated enhanced frontal and transverse compliance while preserving efficient sagittal-plane energy storage and return. The second research objective focuses on translating stiffness variability into a clinically viable solution through manually adjustable sagittal-plane stiffness. A progressive development from the MyFlex-ϵ and MyFlex-ζ prototypes to the Adjustable Pro-Flex Pivot (APP) integrated stiffness modulation into a commercial prosthetic platform. Clinical evaluations confirmed that stiffness preference is highly individual and task-dependent, supporting adjustable solutions beyond conventional weight- and activity-based prescriptions. The refined APP V2 prototype successfully passed ISO 10328 mechanical testing and reached Technology Readiness Level 7, demonstrating structural robustness and near-term clinical applicability. The third research objective introduces MyFlex-θ, a soft-bionic ESR foot incorporating autonomous stiffness adjustment via a convolutional neural network-based activity recognition system. Using embedded inertial sensors, the system classifies locomotor activities in real time and adapts sagittal-plane stiffness accordingly without active propulsion. Experimental results showed high activity classification accuracy and effective stiffness adaptation under real-world conditions. Together, these contributions demonstrate a comprehensive pathway from passive mechanical innovation to intelligent adaptive prosthetic systems, supporting the development of personalized, energy-efficient ESR prosthetic feet capable of adapting to real-world user needs.