Publication: Implementación de algoritmo de scan-matching basado en Clonalg
Loading...
Identifiers
Publication date
2015
Defense date
2015-07-07
Authors
Advisors
Tutors
Journal Title
Journal ISSN
Volume Title
Publisher
Impact
Export
Abstract
La estimación de la posición de robots ha sido una de las ramas de la robótica más estudiada en los últimos años. Es importante para la autonomía de un robot conocer el entorno y posicionarse en él. Para ello se han dado múltiples soluciones al conocido problema del Simultaneous Localization and Mapping (SLAM), que está íntimamente ligado al scan matching o alineación de pares de imágenes.
Para realizar un matching entre imágenes es necesario sacar características del entorno y hacer una asociación de datos correctamente, lo cual es una compleja tarea. Gracias al desarrollo de los sensores en la actualidad se ha conseguido obtener más características del entorno que pueden resultar de gran utilidad para resolver estos problemas, como las características del color.
Por otro lado el progreso y la evolución en el mundo de la inteligencia artificial, IA, ha hecho que se desarrollen potentes algoritmos aplicables a múltiples ámbitos y que aportan soluciones muy satisfactorias y eficientes. Hay una gran variedad de algoritmos, entre ellos están el Differential Evolution (DE) y el Clonal Selection Algorithm (CLONALG). Ambos son utilizados para dar soluciones a múltiples problemas computacionales, y la optimización es uno de ellos.
Este problema de optimización se puede aplicar para realizar el matching o alineamiento entre imágenes. Para ello se debe utilizar una función objetivo o función de coste que se optimizará utilizando un algoritmo. En este trabajo para realizar esta función se va a emplear el CLONALG, un algoritmo evolutivo basado en el principio de selección clonal. Este algoritmo es utilizado en múltiples tareas entre las que destacan el reconocimiento de patrones o la optimización.
Estimating the position of robots has been widely studied during the last decades. To improve the autonomy of a mobilerobot, it is important to obtain an accurate estimate of the robot's location. The Simultaneous Localization and Mapping problem, which is closely related to this task, has multiple solutions and it is closely linked to the scan matching technique. To perform an adequate matching between images or depth scans, it is necessary to have the proper information and it is important to remove unnecessary data, which is a complex task. In the last years, the development of new sensors has provided more detailed information (colour and depth) about images. On the other hand, the constant developments and the evolution in the artificial intelligence research field has resulted inpowerful algorithms that can be applied to multiple tasks. There are multiple methods that can provide satisfactory and efficient solutions to many optimization problems. The evolutionary methods are among them. Two examples are the Differential Evolution and the CLONALG algorithms. The scan matching task can be viewed as an optimization problem. A fitness function can be defined to estimate the quality of the matching process. If the fitness function is implemented in an adequate way, an optimization method can be applied to obtain the solution to the problem. The goal of this work is to implement a CLONALG-based scan matching algorithm. The efficiency will be demonstrated in the experimental results.
Estimating the position of robots has been widely studied during the last decades. To improve the autonomy of a mobilerobot, it is important to obtain an accurate estimate of the robot's location. The Simultaneous Localization and Mapping problem, which is closely related to this task, has multiple solutions and it is closely linked to the scan matching technique. To perform an adequate matching between images or depth scans, it is necessary to have the proper information and it is important to remove unnecessary data, which is a complex task. In the last years, the development of new sensors has provided more detailed information (colour and depth) about images. On the other hand, the constant developments and the evolution in the artificial intelligence research field has resulted inpowerful algorithms that can be applied to multiple tasks. There are multiple methods that can provide satisfactory and efficient solutions to many optimization problems. The evolutionary methods are among them. Two examples are the Differential Evolution and the CLONALG algorithms. The scan matching task can be viewed as an optimization problem. A fitness function can be defined to estimate the quality of the matching process. If the fitness function is implemented in an adequate way, an optimization method can be applied to obtain the solution to the problem. The goal of this work is to implement a CLONALG-based scan matching algorithm. The efficiency will be demonstrated in the experimental results.
Description
Keywords
Robótica, Robots, Proceso de imágenes, Algoritmos evolutivos, Differential Evolution (DE), Clonal Selection Algorithm (CLONALG)