DSpace Repository :: Browsing by Author "Santos, Juan Miguel"

Browsing by Author "Santos, Juan Miguel"

Now showing 1 - 12 of 12

Aprendizaje por refuerzo con opciones y función de refuerzo universal
(2020-12) Bruno Cilla, Diego; Heimann, Matías; Scaglioni, Giuliano; Santos, Juan Miguel
"El objetivo de los algoritmos de aprendizaje por refuerzo es maximizar las recompensas acumuladas a lo largo del tiempo para hallar un comportamiento objetivo. De esta forma, para aprender distintos comportamientos, la variable a cambiar sería la función de refuerzo dada para ese problema. El objetivo de este trabajo es explorar una alternativa en la cual se puedan adquirir distintos comportamientos, manteniendo siempre la misma función de refuerzo, siendo que la variable sea los distintos entornos en los que se realiza el aprendizaje."
Contribución al estudio de la ingeniería inversa de comportamientos emergentes en sistemas multi-agente
(c2012) Parpaglione, María Cristina; Santos, Juan Miguel
"Cada vez que se necesita resolver un problema utilizando un sistema multi-agente se deben responder dos preguntas respecto de los agentes involucrados en el mismo: 1. ¿Qué capacidad de censado debe tener cada uno de ellos? 2. ¿Qué acciones individuales debe tener cada agente para resolver el problema de manera eficiente? Responder cada una de estas preguntas es una tarea difícil cuando se esta resolviendo un problema. Cuando se conocen la capacidad de censado y los comportamientos, el problema a ser resuelto es solamente uno de los posibles comportamientos emergentes del sistema multi-agente en cuestión. El propósito de esta tesis es encontrar un método que permita descubrir cual es la respuesta a cada una de las preguntas anteriores, en orden de obtener un comportamiento emergente dado. Es decir, poder resolver un problema con un grupo de agentes simples con muy baja comunicación entre ellos. Para alcanzar este objetivo, la capacidad de censado de cada agente es modelada con una función parametrizable, encontrando el valor de estos parámetros usando una técnica de optimización, tal como Algoritmos Genéticos. Después de obtener estos valores, cada agente es entrenado utilizando Aprendizaje por Refuerzo para obtener los apropiados comportamientos individuales. Esta tesis propone un método para obtener tanto la capacidad de censado como los comportamientos que debe tener cada agente en un enjambre para alcanzar el comportamiento emergente deseado, dentro del grupo de problemas de formación de patrones. Es decir, resolver esta clase de problema utilizando el paradigma ant-robotic."
Contribución al relevamiento y estado del arte en aprendizaje por refuerzo
(2020-04-24) Emery, Lucas; Santos, Juan Miguel
"Este informe es el resultado del trabajo realizado en el relevamiento, estudio y análisis sobre métodos de aprendizaje por refuerzo y aprendizaje por refuerzo profundo, y fue realizado pensando en constituirse en un recurso para la formación de los interesados en el área."
A control strategy for a tethered follower robot for pulmonary rehabilitation
(2020-12-03) Bianchi, Luciano Gustavo; Buniak, Esteban Alejandro; Ramele, Rodrigo; Santos, Juan Miguel
"Patients that suffer Chronic Obstructive Pulmonary Disease (COPD) undergo a procedure called Pulmonary Rehabilitation that helps them to improve disease prognosis. Pulmonary Rehabilitation consists of different physical exercises and walking activities conducted at medical facilities under supervision of a physical therapist. In order to perform these procedures, patients require oxygen assistance, but the oxygen tank cannot be carried by the patient due to the musculoskeletal atrophy that characterize this pathology and external assistance is required. The assistance to transport the bulky oxygen tank can be provided by a robotic device that follows the patient while performing the physical activities. This work provides an initial study on the controlling mechanism of a differential tethered robot that implements a leader-follower configuration to carry the oxygen tank for these procedures. Two alternative control strategies are proposed. Results on a simulated and on a real prototype confirms the feasibility of the proposed solution."
Discovering sensing capability in multi-agent systems
(2010) Parpaglione, María Cristina; Santos, Juan Miguel
"What should be the sensing capabilities of agents in a Multi-Agent System be to solve a problem efficiently, quickly and economicly? This question often appears when trying to solve a problem using Multi-Agent Systems. This paper introduces a method to find these sensing capabilities in order to solve a given problem. To achieve this, the sensing capability of an agent is modeled by a parametrized function and then Genetic Algorithms are used to find the parameters’ values. The individual behavior of the agents are found with Reinforcement Learning."
EEG waveform analysis of P300 ERP with applications to brain computer interfaces
(2018-11) Ramele, Rodrigo; Villar, Ana Julia; Santos, Juan Miguel
"The Electroencephalography (EEG) is not just a mere clinical tool anymore. It has become the de-facto mobile, portable, non-invasive brain imaging sensor to harness brain information in real time. It is now being used to translate or decode brain signals, to diagnose diseases or to implement Brain Computer Interface (BCI) devices. The automatic decoding is mainly implemented by using quantitative algorithms to detect the cloaked information buried in the signal. However, clinical EEG is based intensively on waveforms and the structure of signal plots. Hence, the purpose of this work is to establish a bridge to fill this gap by reviewing and describing the procedures that have been used to detect patterns in the electroencephalographic waveforms, benchmarking them on a controlled pseudo-real dataset of a P300-Based BCI Speller and verifying their performance on a public dataset of a BCI Competition."
EPOC Emotiv EEG Basics
(2022-10) Ramele, Rodrigo; Villar, Ana Julia; Santos, Juan Miguel
This document provides some basic guidance to start working with the EPOC Emotiv neuroheadset device and describes how to use it to perform basic Brain-Computer Interface (BCI) research. A brief tutorial on how to set up the device, from its electrophysiological point of view, as well as a description and practical code to perform some basic analysis, is explained. A basic experiment is introduced to detect one of the oldest and, indeed, quite still valuable electrophysiological correlate, visual occipital alpha waves, or Berger Rhythm. An additional experiment is expounded where the power spectrum of alpha waves is reduced when a subject is affected by background cognitive disturbances. This document also briefs about the extraction of information by using the EPOC Emotiv library and also with python Emokit package. This report presents a basic guide on how to use EEGLAB + MATLAB, as well as python stack to perform the neurophysiological analysis. Finally, a basic analysis on different feature extraction and classification methods is provided.
Equilivest: A robotic vest to aid in post-stroke dynamic balance rehabilitation
(2023-01) Paviotti, Franco; Buniak, Esteban; Ramele, Rodrigo; Freixes, Orestes; Santos, Juan Miguel
Brain stroke is a devastating medical condition, that affects world population and is the main cause of disabilities worldwide. Disabilities related to stroke can affect motor pathways, and may lead to several motor function disorders. One important aspect of motor function is balance which is the ability to control the body’s center of mass inside the base support provided by the lower limb. Stroke can affect dynamic balance as well, which is manifested while walking, impairing autonomy and independence, important factors in Activities of Daily Living (ADL) particularly for young patients.
Esquema actor-crítico en aprendizaje por refuerzo con espacios continuos
(2020-12-21) Delgado, Francisco; Santos, Juan Miguel
"El siguiente informe es el resultado del trabajo de relevamiento, estudio y desarrollo de métodos de aprendizaje por refuerzo principalmente enfocados al esquema Actor-Crítico. A lo largo del texto, se introducen los conceptos base de la teoría y se plasman los fundamentos y teoría no solo de el esquema Actor-Crítico, sino también su predecesor REINFORCE, donde se introduce el concepto de optimización de políticas. El esquema Actor-Crítico es instanciado mediante la implementación de Proximal Policy Optimization (PPO) el cual utiliza redes neuronales artificiales(RNA) como estimador de función no lineal. Para el ajuste de los pesos de las RNAs se evaluaron tres variantes del gradiente estocástico las cuales incluyen Momentum, RMSProp y ADAM. Se diseñaron una serie de experimentos con el propósito de comparar el esquema Actor-Crítico con REINFORCE y sus resultados permiten establecer las diferencias de eficiencia entre ambos. Estos experimentos se hacen en base al problema del péndulo invertido en un ambiente dinámico. Finalmente, para demostrar la robustez y flexibilidad de los métodos de Actor-Crítico toma como caso de estudio un problema de control complejo en donde un agente debe aprender a caminar. Dicho agente es una criatura artificial semejante a una hormiga de cuatro patas, con dos rotores por pata que aplican torque sobre las mismas. La evaluación del desempeño del agente se realiza midiendo la longitud de la trayectoria recorrida hacia un objetivo en una cantidad de tiempo."
Histogram of gradient orientations of EEG signal plots for brain computer interfaces
(2018) Ramele, Rodrigo; Santos, Juan Miguel; Villar, Ana Julia
"Brain Computer Interface (BCI) or Brain Machine Interfaces (BMI), has proved the feasibility of a distinct non-biological communication channel to transmit information from the Central Nervous System (CNS) to a computer device. Promising success has been achieved with invasive BCI, though biocompatibilities issues and the complexity and risks of surgical procedures are the main drive to enhance current non-invasive technologies. Electroencephalography (EEG) is the most widespread method to gather information from the CNS in a non-invasive way. Clinical EEG has traditionally focused on temporal waveforms, but signal analysis methods which follow this path have been neglected in BCI research. This Thesis proposes a method and framework to analyze the waveform, the shape of the EEG signal, using the histogram of gradient orientations, a fruitful technique from Computer Vision which is used to characterize image local features. Inspiration comes from what traditionally electroencephalographers have been doing for almost a century: visually inspecting raw EEG signal plots."
Histogram of gradient orientations of signal plots applied to P300 detection
(2019-07) Ramele, Rodrigo; Villar, Ana Julia; Santos, Juan Miguel
"The analysis of Electroencephalographic (EEG) signals is of ulterior importance to aid in the diagnosis of mental disease and to increase our understanding of the brain. Traditionally, clinical EEG has been analyzed in terms of temporal waveforms, looking at rhythms in spontaneous activity, subjectively identifying troughs and peaks in Event-Related Potentials (ERP), or by studying graphoelements in pathological sleep stages. Additionally, the discipline of Brain Computer Interfaces (BCI) requires new methods to decode patterns from non-invasive EEG signals. This field is developing alternative communication pathways to transmit volitional information from the Central Nervous System. The technology could potentially enhance the quality of life of patients affected by neurodegenerative disorders and other mental illness. This work mimics what electroencephalographers have been doing clinically, visually inspecting, and categorizing phenomena within the EEG by the extraction of features from images of signal plots. These features are constructed based on the calculation of histograms of oriented gradients from pixels around the signal plot. It aims to provide a new objective framework to analyze, characterize and classify EEG signal waveforms. The feasibility of the method is outlined by detecting the P300, an ERP elicited by the oddball paradigm of rare events, and implementing an offline P300-based BCI Speller. The validity of the proposal is shown by offline processing a public dataset of Amyotrophic Lateral Sclerosis (ALS) patients and an own dataset of healthy subjects."
Training a gaming agent on brainwaves
(2020-12-07) Bartolomé, Francisco; Moreno, Juan; Navas, Natalia; Vitali, José; Ramele, Rodrigo; Santos, Juan Miguel
"Error-related potential (ErrP) are a particular type of Event-Related Potential (ERP) elicited by a person attending a recognizable error. These Electroencephalographic (EEG) signals can be used to train a gaming agent by a Reinforcement Learning (RL) algorithm to learn an optimal policy. The experimental process consists of an observational human critic (OHC) observing a simple game scenario while their brain signals are captured. The game consists of a grid, where a blue spot has to reach a desired target in the fewest amount of steps. Results show that there is an effective transfer of information and that the agent successfully learns to solve the game efficiently, from the initial 97 steps on average required to reach the target to the optimal number of 8 steps. Our results are expressed in threefold: (i) the mechanics of a simple grid-based game that can elicit the ErrP signal component, (ii) the verification that the reward function only penalizes wrong steps, which means that type II error (not properly identifying a wrong movement) does not affect significantly the agent learning process; (iii) collaborative rewards from multiple observational human critics can be used to train the algorithm effectively and can compensate low classification accuracies and a limited scope of transfer learning schemes."

Browsing by Author "Santos, Juan Miguel"

Results Per Page

Sort Options