Planning and estimation algorithms for human-like grasping

• Autores: David Álvarez Sánchez
• Directores de la Tesis: Luis Enrique Moreno Lorente (dir. tes.), Luis Santiago Garrido Bullón (codir. tes.)
• Lectura: En la Universidad Carlos III de Madrid ( España ) en 2016
• Idioma: español
• Tribunal Calificador de la Tesis: Carlos Balaguer Bernaldo de Quirós (presid.), Raúl Suárez Feijóo (secret.), Pedro Lima (voc.)
• Materias:
  • Matemáticas
    • Ciencia de los ordenadores
      • Inteligencia artificial
    • Investigación operativa
      • Planificación
• Enlaces
  • Tesis en acceso abierto en: e-Archivo
• Resumen

  • The use of robots in human like environment requires them to be able to sense and model unstructured environments. Their success will depend on their versatility for interacting with the environment. This interaction often includes manipulation of objects for accomplishing common daily tasks. Therefore, robots need to sense, understand, plan and perform; and this has to be a continuous loop.

    This thesis presents a framework which covers most of the phases encountered in a common manipulation pipeline. First, it is shown how to use Fast Marching Squared algorithm and a leader-followers schema to control a formation of robots, simplifying and high dimensional path-planning problem. Then, for the motion planning part of the grasping action, this concept is recycled from the robot formation techniques to be applied to the control of complex hand-arm systems. Besides, the necessary techniques for environment modelling and grasp synthesis are also presented.

    Finally, under the assumption that the grasping actions may not always result as it was previously planned, a Bayesian based state-estimation process is introduced to estimate the final in-hand object pose after a grasping action is done, based on the measurements of propioceptive and tactile sensors.