Development of a Control System for a Biomechanical Prosthetic

This work details the development of a control system for the MAKI biomechanical hand prosthetic, developed by the Biomedical Engineering Research Group (GIIB). For the control of the prosthetic, the use of hierarchical control, consisting of 3 levels: high, medium, and low-level control is proposed. In the high-level control, a neural network is developed to classify movements, with the Electromyography (EMG) signals as input and the movement required by the user as output. In the medium-level control, a finite state machine is developed which is in charge of guaranteeing the coherence of the movements and eliminating possible errors that the neural network may have; finally, as a low-level control, positioning control is implemented in this way, guaranteeing that the prosthetic can perform the established movements. In this research, it has been determined that using a state machine eliminates some errors that can be generated when changing from one state to another. In the same way, an essential aspect of a classifier is segmentation; the greater the segmentation, the lower the error in the classifier.