Método de evaluación dinámica de planes en sistemas inteligentes autónomos.
Files
Date
2015-06-26
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Facultad Regional Buenos Aires
Abstract
La característica principal de los sistemas inteligentes autónomos es que son capaces de auto proponerse planes, de ejecutarlos y de retroalimentar su base de conocimiento a partir de la información que extraen del entorno. En esta tesis se presenta una revisión de los métodos de aprendizaje y planificación de dichos sistemas, para luego centrar la investigación sobre la arquitectura LOPE. Aunque publicaciones posteriores implementaron modificaciones y extensiones que lograron mejorar su rendimiento, se han identificado ciertos aspectos del modelo que aún no han sido abordados. Por tal motivo, se proponen mejoras para ser aplicadas dentro de los módulos de planificación y aprendizaje, como también refinamientos al proceso de control y ejecución. Además, se elabora un indicador que permite una evaluación integral de la arquitectura y un análisis comparativo de los resultados alcanzados con las distintas mejoras aplicadas, en comparación con el diseño original.
The main feature of autonomous intelligent systems is their ability to self-propose plans, to execute them and to provide feedback to its knowledge base from the information that they extract from the environment. In this thesis, we present a review of the learning and planning methods of the above mentioned systems, in order to focus the research on the LOPE architecture later on. Although subsequent researches implemented modifications and extensions which got to improve its performance, we have identified certain aspects of the model that have not yet been addressed. Therefore, several improvements have been proposed in order to be applied within the planning and learning modules, as well as refinements to the control and execution process. Furthermore, we elaborate an indicator that allows us to get a global evaluation of the architecture and to perform a comparative analysis between the results obtained through each of the implemented enhancements, compared to the original design.
The main feature of autonomous intelligent systems is their ability to self-propose plans, to execute them and to provide feedback to its knowledge base from the information that they extract from the environment. In this thesis, we present a review of the learning and planning methods of the above mentioned systems, in order to focus the research on the LOPE architecture later on. Although subsequent researches implemented modifications and extensions which got to improve its performance, we have identified certain aspects of the model that have not yet been addressed. Therefore, several improvements have been proposed in order to be applied within the planning and learning modules, as well as refinements to the control and execution process. Furthermore, we elaborate an indicator that allows us to get a global evaluation of the architecture and to perform a comparative analysis between the results obtained through each of the implemented enhancements, compared to the original design.
Description
Keywords
Ingeniería en Sistemas de Información, Sistemas inteligentes autónomos, Aprendizaje por interacción con el entorno, Exploración, Planificación, Formación y revisión de teorías, Aprendizaje por refuerzo, Necesidades y motivación, Autonomous intelligent systems, Environment-interaction based learning, Exploration, Planning, Theory creation and revision, Reinforcement learning, Needs and motivation
Citation
Endorsement
Review
Supplemented By
Referenced By
Creative Commons license
Except where otherwised noted, this item's license is described as info:eu-repo/semantics/openAccess