Publication: Evolución de árboles en el problema del wall following robot
Loading...
Identifiers
Publication date
2012-03-13
Defense date
2012-03-12
Authors
Advisors
Tutors
Journal Title
Journal ISSN
Volume Title
Publisher
Impact
Export
Abstract
Este proyecto trata la hipótesis de que existe una relación entre forma arbórea de los individuos y su adaptación a las soluciones, permitiendo estimar el fitness de los individuos en función de su forma arbórea. Para ello se ha implementado el problema del wall following robot, según lo describió John R. Koza en su publicación “Evolution of Subsumption Using Genetic Programming”. Finalmente se comentarán las conclusiones extraídas a partir de los resultados obtenidos del motor de programación genética ProGen. Estos resultados constarán de una comparación de rendimiento entre la programación genética en árbol y la programación genética tradicional, evaluando tanto los resultados brutos, como el coste empleado para extraerlos. Por último se implementó una herramienta de visualización de resultados, que permite poder interpretar de forma intuitiva los experimentos. El movimiento de los robots se representa como una línea azul dentro de la habitación de paredes rojas como se muestra en las distintas capturas de pantalla de este documento. _________________________________________________________________________________________________________________________
This Project deals with the hypothesis that there is a relationship between the individual’s tree form and their way to find a solution, allowing us to estimate the fitness value of the individuals based on their tree form. In order to fulfill the hypothesis this Project has developed the Wall Fallowing Robot problem as described by John R. Koza in his paper “Evolution of Subsumption Using Genetic Programming”. The ProGen genetic programming engine was used to achieve this and obtain the results discussed in the conclusions chapter of this memo. Finally, a viewer was developed to be able to easily show the results of each experiment. In this viewer the path followed by the robots is represented by a blue line inside of a red square room. Screen captures of this viewer have been included in this document.
This Project deals with the hypothesis that there is a relationship between the individual’s tree form and their way to find a solution, allowing us to estimate the fitness value of the individuals based on their tree form. In order to fulfill the hypothesis this Project has developed the Wall Fallowing Robot problem as described by John R. Koza in his paper “Evolution of Subsumption Using Genetic Programming”. The ProGen genetic programming engine was used to achieve this and obtain the results discussed in the conclusions chapter of this memo. Finally, a viewer was developed to be able to easily show the results of each experiment. In this viewer the path followed by the robots is represented by a blue line inside of a red square room. Screen captures of this viewer have been included in this document.
Description
Keywords
Robótica, Algoritmos genéticos, Inteligencia artificial