Biomechanics and Ergonomics

This project aims to contribute to a better knowledge of the human grip from the synergy between artificial intelligence techniques widely used in the field of robotics and techniques dimensionality reduction well known in the field of biomechanics of the human hand.  From this synergy emerges the main objective of the project: determining habitual postural patterns present in activities of daily living that could orient the control of a virtual prosthetic hand with high mobility.
To date a variety of robotic and prosthetic hands have been developed, some to achieve a stable grip and dexterity similar to that of the human hand and others for a better cosmetic appearance and handling anthropomorphism. However, current prosthetic hand are very simple from a biomechanical point of view. The problem lies in the difficulty of establishing adequate communication between the user and the prosthesis. There is no simple way to achieve the interaction between the right hand amputated and multiple GDL prostheses, such as those developed in recent decades, as it requires either too independent control signals or an intelligent controller. Trends in the control of robotic and prosthetic hands indicate a novelty in this area could be achieved through observation and imitation of the natural biomechanical behavior, based on a space of reduced dimensionality.
The paper proposes the use of principal component analysis to reduce the dimensionality of the problem of control in conjunction with artificial neural networks, that will generate automatically gripping positions for a natural-looking virtual hand high mobility in clamp and cylindrical grips of cylindrical objects of different diameters. This way the minimum number of control inputs necessary can be establish to get an advanced prosthetic hand which can perform activities of daily life based on identified patterns and determine their ability to control in real time, through the use of Artificial Neural Networks. This approach represents a contribution to the study of prosthetic control, since the use of artificial neural networks is a methodology for estimating and monitoring seldom used in this area and has great advantages of implementation and execution in real time.
To achieve the main objective of the project five specific objectives have been set. The first is to obtain a database of different gripping positions subjects performing grips with cylindrical objects of different diameters. The second addresses the reduction of the dimensionality of the grips made by principal component analysis, determining postural synergies produced during the grip of the human hand. The third involves the implementation of two artificial neural networks capable of automating the estimation of the positions of grip from postural synergies identified to grips studied. The fourth objective involves the development of a virtual hand model of 25 dof for simulating grip positions obtained through the use of open source 3D modeling and simulation. Finally, the fifth objective will involve implementation and evaluation by simulating different control strategies for the position of the virtual hand generated taking as control inputs the obtained postural synergies.
The proposed research may be of great interest in various applications such as: improving design of prosthetic bionic arms, developing efficient robotic hands for household or industrial applications, development of virtual reality haptic devices, improved design products or manipulated by people and assisting surgeons in surgical planning tools on hand.