Robot grasping in 3D through efficient haptic exploration with unscented Bayesian optimization and collision penalty

06/08/2018

Robust grasping of arbitrary objects is a major, and still unsolved problem in robotics. Global 3D information about the object can be obtained from previous knowledge (e.g., accurate models of known objects or approximate models of familiar objects) or real-time sensing (e.g., partial point clouds of unknown objects) and can be used to identify good potential grasps; however, due to modeling inaccuracies and sensing limitations, local exploration is often needed to refine such grasps and to successfully apply them to the objects in the real world. Notably, the recently proposed unscented Bayesian optimization technique can make such exploration safer by favoring the selection of grasps that are robust to uncertainty in the input space (e.g., inaccuracies in the execution of the grasp). By finding a safer grasp, we assure that a slight error in grasp position due to execution noise won't cause a dramatic decrease of the grasp quality. Extending the previous work done on 2D optimization, in this thesis we propose a 3D haptic exploration strategy that combines unscented Bayesian optimization with a novel collision penalty heuristic to find safe 3D grasps in a very efficient way. By expanding the search-space from 2D to 3D we are able to find better grasps and the introduction of the collision penalty heuristic allows to do so without having to increase the number of exploration steps, therefore battling the curse of dimensionality.

Authors in the community:

• 

  J

  João Pedro Morais Castanheira

  ist177206

Supervisors of this institution:

• 

  A

  Alexandre José Malheiro Bernardino

  ist13761

electrical-engineering-electronic-engineering-information-engineering - Electrical engineering, electronic engineering, information engineering

eng - English

Instituto Superior Técnico