Study On Autonomous Grasp Planning For Multifingered Hands

As an end-effector of a robot manipulator, the hand realizes grasping and dexterous manipulating for different objects through imitating human hands, which has attracted serious attentions in robot community. So, the grasp planning algorithm of multifingered hands has become the focus of this area. In this dissertation, the problem of the grasp planning of multifingered hands is discussed.Firstly, a model of multifingered hands is provided, and its kinematics equations are established. Then, the forward and inverse kinematics of multifingered hands in joint space and the redundancy problem of the single finger are analyzed, which are followed by correlative simulation tests.Secondly, an autonomous grasp planning method is studied in consideration of the characters of the objects and the grasp tasks, which includes: (1) Grasp mode planning algorithm. A knowledge base of grasp mode is constructed. Then the classification problem is studied using BPNN and RBFNN respectively. Simulation results show that the RBFNN has better classification performance. (2) Grasp force planning algorithm. By the SUMT technique, this dissertation provided a minimum grasp force algorithm to solve the multifingered grasp force problem. Simulation tests illustrate that the algorithm has good performance. (3) The optimal grasp planning of multifingered hands is discussed based on approach distance. Then, a visual platform is developed by VC/OpenGL to demonstrate the grasping planning problem vividly. Finally, a novel concept named the modular self-reconfigurable robot hand is presented according to the Holonic grasp and the modular self-reconfigurable idea.This kind of robot hand has the autonomous capacity to adjust its configuration for adaptation to variant grasp tasks. Hence, it can be used in more complicated situations. The corresponding optimal topology configuration transition problem is studied based on the generation algorithm, and the process is also simulated.