GRASP PLANNING FOR ROBOTIC MULTI-FINGERED HANDS

Grasp planning is one of the key issues for dexterous robotic hands to accomplish desired tasks.A promising approach is to teach the robot hands by human operator.Due to the technical limitation of human machine interface,the grasp configuration learning from the human hand needs to be improved.This paper presents a method to optimize the grasp from the initial configuration provided by the human hand.A quality measure based on joint motion isotropy is defined to evaluate grasp configurations.Human grasp experience is used to provide a desired initial configuration.The position and orientation of the palm is optimized based on the quality measure.The process of optimization is simulated in a virtual environment and the resulting optimal grasps illustrate the validity of the method.