Publication: Evaluación de técnicas evolutivas en el cálculo de la cinemática inversa de robots manipuladores
Loading...
Identifiers
Publication date
2012-07
Defense date
2012-07-19
Authors
Advisors
Tutors
Journal Title
Journal ISSN
Volume Title
Publisher
Impact
Export
Abstract
En este proyecto, se examina el rendimiento de un algoritmo de búsqueda directa a partir del problema de cinemática inversa de robots manipuladores con diferentes grados de libertad. Para ello se plantea el siguiente problema de optimización: conocida la posición y orientación deseada en el extremo nal de un robot manipulador, y su función de cinemática directa, encontrar un posible vector de con guración articular óptimo dentro del espacio de trabajo del robot, de manera que, los errores de posición y orientación son mínimos. Un enfoque evolutivo, llamado Di erential Evolution, es el utilizado para resolver el anterior problema de optimización con restricciones no lineales. Di erential Evolution es un método de búsqueda directa estocástico basado en un proceso iterativo de mutación algebraica, cruce y selección sobre una población de soluciones candidatas. Basada en la experimentación empírica, este estudio pretende mostrar la in uencia de los valores de los parámetros de control (tamaño de la población, factor de escala de mutación, etc.) en el comportamiento del algoritmo para el problema de inversión cinemática. Finalmente, se introduce un operador de búsqueda local para mejorar la velocidad de convergencia del algoritmo. ______________________________________________________________________________________________________________________
In this dissertation, the performance of a direct search algorithm is examined for the inverse kinematics problem of robot manipulators with di erent degrees of freedom. The following optimization model has been considered: given a desired end-e ector position and orientation, and the forward kinematics function of a robot manipulator, nd a feasible optimal joint con- guration vector within the robot's workspace, such that, the position and orientation errors are minimal. An evolutionary-based approach, called Di erential Evolution, is used to solve the above nonlinear constrained optimization problem. Di erential Evolution is a stochastic direct search method based on an iterative process of algebraic mutation, crossover, and selection over a population of candidate solutions. Based on empirical experimentation, the study aims to reveal the in uence of control parameters values (population size, mutation scale factor, etc.) on the algorithm behaviour for the kinematic inversion problem. Finally, a local search operator is introduced to improve the convergence speed of the algorithm.
In this dissertation, the performance of a direct search algorithm is examined for the inverse kinematics problem of robot manipulators with di erent degrees of freedom. The following optimization model has been considered: given a desired end-e ector position and orientation, and the forward kinematics function of a robot manipulator, nd a feasible optimal joint con- guration vector within the robot's workspace, such that, the position and orientation errors are minimal. An evolutionary-based approach, called Di erential Evolution, is used to solve the above nonlinear constrained optimization problem. Di erential Evolution is a stochastic direct search method based on an iterative process of algebraic mutation, crossover, and selection over a population of candidate solutions. Based on empirical experimentation, the study aims to reveal the in uence of control parameters values (population size, mutation scale factor, etc.) on the algorithm behaviour for the kinematic inversion problem. Finally, a local search operator is introduced to improve the convergence speed of the algorithm.
Description
Keywords
Robótica, Optimización matemática, Algoritmos genéticos