Enhancing Upper-Limb Prostheses via Multisensing and Semi-Autonomous Sensorimotor Control Strategy

Despite remarkable technological advancements, current upper-limb prostheses still face critical limitations that hinder their effective use in real-world scenarios. These include restricted gesture decoding capabilities, insufficient environmental sensing, high cognitive demands, and poor adaptability to dynamic contexts. This thesis addresses these challenges by proposing and validating innovative semi-autonomous sensorimotor prosthetic systems that integrate advanced control strategies with multisensory information to enhance human-device interaction. Taking inspiration from human sensorimotor behaviour during grasping and manipulation tasks, the proposed framework combines muscular activity decoding with environmental perception to develop control strategies aimed at improving prosthesis functionality. Going into the details of the proposed activities to reach this goal, a novel wearable sensing device based on displacement myography was explored and validated as an alternative to traditional electromyography for recording muscular activity. It demonstrated high accuracy and reliability across a wide range of 33 gestures, including hand gestures, force levels, and varying degrees of finger aperture. Although displacement myography consistently outperformed electromyography in overall performance, a task-specific analysis revealed complementary strengths within each modality: the former was particularly more effective at recognizing different degrees of finger aperture, while the latter showed relatively better performance in distinguishing between force levels. Nonetheless, displacement myography remained superior even in the recognition of force levels, confirming its overall advantage. Such technology proved to be a reliable and user-friendly interface for accurate motion decoding; consequently, it paves the way for a potential application in the field of prosthetic control. Regarding the environmental perception, commercially available force and temperature sensors were evaluated and selected based on cost-effectiveness, compactness, compatibility with prosthetic integration and functional alignment with physiological characteristics of the human hand. These sensors were systematically tested under realistic conditions to ensure reliable performance in everyday use and integrated into a prosthetic hand. The acquired multisensory data, comprising both muscular decoding and environmental perception, were leveraged to develop two semi-autonomous sensorimotor control strategies able to improve prosthesis performance. The first strategy utilized force and temperature information to enhance grasp stability by managing slip events and regulating grip force, while also protecting the user and the device from harmful thermal conditions. The second strategy dynamically adjusted wrist velocity based on interaction forces, improving functional performance while reducing compensatory movements without increasing cognitive load. Both strategies were rigorously evaluated through simulations, benchtop tests, and user trials, including an amputee participant. In conclusion, this work demonstrates the potential of integrating user intention decoding with tactile environmental sensing to overcome significant limitations of current prosthetic systems. The proposed multisensing framework is the basis for developing control strategies that pave the way for more intuitive, adaptive, and intelligent prostheses, offering new perspectives for the development of next-generation assistive technologies.