| Abstract: |
The use of the same end-effector to grasp an object in various orientations performing different tasks dramatically increases its capabilities. This can be achieved by alternating grasp configurations of the object with respect to the task to be done. This is known as Regrasping. A basic component of object manipulation is the need to regrasp the object at a new configuration. However, current regrasping methodologies work only with highly redundant (and hence expensive) hand architectures, and require overly sophisticated sensory feedback. We present a novel motion planning computation algorithm to perform a rapid and optimal dynamic regrasp manipulation with a fully actuated robotic arm and a non-dexterous end-effector. The algorithm is a semi-stochastic method which parameterizes the whole regrasping motion as a single vector. The constraints of the problem are formulized in terms of time and the parameterization vector. The algorithm generates numerous parameterization vectors and check their feasibility based on the constraints. This paper presents the novel algorithm and the current stage of the research. The desired outcome of the research is also presented. |