Adaptive adjustment of regrasping for object's posture error

This paper proposes adaptive adjustment methods of a preplanned regrasping strategy according to position and orientation errors of a grasping object. We had already derived a generation algorithm for the regrasping strategy by evolutionary programming. However position and orientation errors of a grasping object from those in numerical simulation are experimentally inevitable since there are many indeterminable factors such as slip contact between a fingertip and an object surface, imbalance of grasping forces, and so on. Those errors are getting bigger according to the task progress, and it is finally difficult to recover those errors due to a limitation of hand's work space and so on. We therefore propose some adaptive adjustment methods that can change an grasping point or intermediate posture of a grasping object so that it could continue to manipulate an object against larger errors. We investigate the allowable error range of the proposed methods in numerical simulations and confirm its reality by experiment.