Dynamic Grasp Synthesis of Simple Convex Shapes

The general grasp planning problem has been studied in the academic community for decades. Developing efficient and versatile grasping systems has great implications in automating tasks that require dexterous and compliant manipulation. Traditional approaches tend to favor optimality over efficiency, producing grasping systems that are slow. I propose a new strategy that combines successful probabilistic approaches, object model abstraction schemes and force-closure constraint approximations into a single unifying algorithm. It is my hypothesis that through combining successful techniques already published in the literature, a more efficient strategy can be developed. The proposed algorithm will be tested in the dynamic simulation environment GraspIt! and then tested on the physical Barrett Hand and WAM arm