A memetic approach to the inverse kinematics problem for robotic applications
- Autores: Carla Elena González Uzcátegui
- Directores de la Tesis: Dolores Blanco Rojas (dir. tes.)
- Lectura: En la Universidad Carlos III de Madrid ( España ) en 2014
- Idioma: inglés
- Tribunal Calificador de la Tesis: Pedro Lima Lima (presid.), Mohamed Abderrahim Fichouche (secret.), Manuel Ferre Pérez (voc.), Cecilia García Cena (voc.)
- Enlaces
- Tesis en acceso abierto en: e-Archivo
- Resumen
The inverse kinematics problem of an articulated robot system refers to computing the joint configuration that places the end-effector at a given position and orientation. To overcome the numerical instability of the Jacobian-based algorithms around singular joint configurations, the inverse kinematics is formulated as a constrained minimization problem in the configuration space of the robot. In previous works this problem has been solved for redundant and non-redundant robots using evolutionary-based algorithms. However, despite the flexibility and accuracy of the direct search approach of evolutionary algorithms, these algorithms are not suitable for most robot applications given their low convergence speed rate and the high computational cost of their population-based approach. In this thesis, we propose a memetic variant of the Differential Evolution (DE) algorithm to increase its convergence speed on the kinematics inversion problem of articulated robot systems. With the aim to yield an efficient trade-off between exploration and exploitation of the search space, the memetic approach combines the global search scheme of the standard DE with an independent local search mechanisms, called discarding. The proposed scheme is tested on a simulation environment for different benchmark serial robot manipulators and anthropomorphic robot hands. Results show that the memetic differential evolution is able to find solutions with high accuracy in less generations than the original DE. -----------------------------------------------------------
