Politecnico di Torino (logo)

EMG-Driven​ ​control​ ​in the​ ​Human​ ​Lower Prosthesis

Maximilian Maniacco

EMG-Driven​ ​control​ ​in the​ ​Human​ ​Lower Prosthesis.

Rel. Marcello Chiaberge. Politecnico di Torino, Corso di laurea magistrale in Mechatronic Engineering (Ingegneria Meccatronica), 2019

PDF (Tesi_di_laurea) - Tesi
Licenza: Creative Commons Attribution Non-commercial No Derivatives.

Download (7MB) | Preview

Recent technological improvements in the field of neurorobotics have led to developments of numerous wearable robotic assistive devices in the field of rehabilitation. Such new prosthetic and orthotic devices should ideally recognize a users intention to act and aid with different environmental interactions. While the mechanical design of such devices remains a demanding challenge, it is equally important to address and investigate novel control strategies that allow the control of these devices in an intuitive and seamless way. In fact, the objective of novel control paradigms in prostheses and orthoses is to recognize effectively the user intention through bio-signals those generate spontaneously during movement, like electromyography. Even though myoelectric prostheses are nowadays a common reality for upper limb amputees, none of the available lower limb prosthesis on the market provide EMG-driven control. The aim of this work is therefore the definition of a novel approach for myoelectric control in lower limb prosthesis capable of overcome state-of-the-art controller limitation. In fact, since modern lower limb prostheses can generate net positive mechanical work, they also have the capability to restore natural gait in amputees. Configuration of these devices imposes a new control layout with respect to conventional prostheses. The new control layout driven from user intention to move should consider peculiar differences between rhythmic and the voluntary movements, in order to combine and optimize the control strategy. To successfully accomplish this task, a novel machine learning based approach for gait-cycle classification and impedance law controller is derived. The first layer of the proposed control framework is implemented with supervised machine learning classifiers for optimal walking phase identification. For each identified phase impedance control law parameters are subsequently tuned automatically through gaussian regression and applied to the knee and ankle joints of the prostheses. Both classification and regression machine learning models are investigated through input signals and parameters combination. Performance of the controller is evaluated through online simulations of the controller: minimization of the error between the predicted control output and the measured physiological joint torque is used to assess the different proposed approaches. The results show that inclusion of EMG-signals in the control framework do not increase gait-cycle classification accuracy with respect to traditional sensors. Instead, the presence of EMG sensors was demonstrated to be favorable during motor task transitions, such as start walking or standing up. In fact, the extension of the classifier with an additional layer to distinguish between standing or walking condition performed better with EMG signals. This demonstrate the importance of introducing EMG signals for non-rhythmic movement control.

Relators: Marcello Chiaberge
Academic year: 2018/19
Publication type: Electronic
Number of Pages: 90
Corso di laurea: Corso di laurea magistrale in Mechatronic Engineering (Ingegneria Meccatronica)
Classe di laurea: New organization > Master science > LM-25 - AUTOMATION ENGINEERING
Aziende collaboratrici: FONDAZIONE IIT
URI: http://webthesis.biblio.polito.it/id/eprint/10917
Modify record (reserved for operators) Modify record (reserved for operators)