Control and Rehabilitation of Biomechanical Systems using Reinforcement Learning
Visualitza/Obre
Estadístiques de LA Referencia / Recolecta
Inclou dades d'ús des de 2022
Cita com:
hdl:2117/355796
Realitzat a/ambCentre de Recerca en Enginyeria Biomedica (CREB)
Tipus de documentProjecte Final de Màster Oficial
Data2021-10-15
Condicions d'accésAccés obert
Llevat que s'hi indiqui el contrari, els
continguts d'aquesta obra estan subjectes a la llicència de Creative Commons
:
Reconeixement-NoComercial-SenseObraDerivada 3.0 Espanya
Abstract
The human nervous system is a complex neural network that is capable of learning a wide variety of cognitive and motor skills. The brain processes and interprets the electrical signals that come from the senses, and generates a response signal that activates and controls our muscles. The optimal pathways for a wide variety of responses are continuously being learned and improved by the nervous system since birth, which makes the modeling of the control system of the human body quite challenging. Moreover, the rehabilitation process, where the brain tries to find new pathways to bypass the damaged ones, is not yet fully understood. Reinforcement learning (RL) techniques are very effective for controlling the motion of mechanical systems. For instance, control systems based on artificial neural networks can learn optimal control strategies in both simulated environments and the real world. RL algorithms can also adapt and learn any changes in the system, which makes them very interesting for rehabilitation. The main goal of the project is to understand the changes that occur in our neural system during learning and rehabilitation using RL algorithms based on artificial neural networks, and ultimately assess and improve patient rehabilitation techniques. In addition to developing control strategies for biomechanical models of the human body to perform different tasks (walking, jumping, running, etc.), as well as analyzing both physiological and non-physiological cost (or reward) functions and the possible relations between them. The student will have to understand how the biomechanics of the human body is modeled and simulated, and will perform simulations of different tasks with a human body model using OpenSim in Python. Optimization of the learning process will also be explored, using for instance solutions of the classical optimal control problem to bootstrap the RL algorithm and reduce training time.
MatèriesArtificial arms -- Design and construction -- Computer simulation, Artificial intelligence -- Medical applications -- Evaluation -- Mathematical models, Braços artificials -- Disseny i construcció -- Simulació per ordinador, Intel·ligència artificial -- Aplicacions a la medicina -- Avaluació -- Models matemàtics
TitulacióMÀSTER UNIVERSITARI EN ENGINYERIA INDUSTRIAL (Pla 2014)
Col·leccions
Fitxers | Descripció | Mida | Format | Visualitza |
---|---|---|---|---|
tfm-qmb.pdf | 16,02Mb | Visualitza/Obre |