Grasp Planning And Experiment Of Humanoid Multi-fingered Dexterous Hands

Humanoid Multi-fingered Dexterous hands are a kind of manipulation of robots.Because of their special structure and dexterous manipulating method,they can provide much convenience for humankind in production and living.Due to the large degrees of freedom and various grasping modes of multifingered hands,it is complex to plan the hands' grasping.In order to make it possible for an-finger dexterous hand to grasp an-polygon object after acquiring the vertex information of the object,and to grasp it with a minimum force avoiding breaking the object,this paper provides a grasp planning algorithm,which is also able to resist known external disturbance.Experiment is performed for verification.The main contents are as follows:A BarrettHand is analyzed for mathematical model,using D-H method to calculate displacement transformation matrix,based on which both forward and inverse kinematic are analyzed and simulated.Dynamic function is also provided using Lagrangian function,in order to find out the relationship between each joint's torque and angular displacement,angular velocity and angular acceleration.According to the characteristics of multifingered hands,a grasp planning algorithm is provided.Based on the shape of an object and the number of fingers of the hand,a problem description is presented and simplified.Relying on the theory of characteristic plane,force-closure,friction cone and related theories,an algorithm calculating the positions is put forward to find out an initial grasping posions;afterwards,according to a grasping performance index named the maximum internally tangent sphere,aiming at maximum stability,the most stable grasp positions are determined by an iteration.Based on these obtained positions and the necessity of the minimum force,an algorithm calculating the minimum applied force by each finger of an-finger hand is provided.Hard finger model is chosen and the grasp matrix is obtained.Applied force is divided into grasp force and manipulation force,whose definition and properties are given.Similarly focus on force-closure and friction cone,a method of calculating applied force produced.Based on external penalty function,the minimum force is found out.This method also works well when external disturbance occurs.Use V-REP for simulation,verifying forward kinematics,inverse kinematics and dynamic function.In addition,the whole system is setup and simulated to show the algorithm's validity and safety.An experiment is established,with host computer,UR and BarrettHand.Use Harris corner detection to get the vertex information of the objects,grasping two groups of objectives.The result shows that this algorithm has a high success rate for grasping,and is able to grasp different shapes of objects,also realizing the function of resisting external force wrench.A series of measurement experiments are processed to get force data during the grasping.By comparing the experiment results and calculation results,it is verified that the grasp planning algorithm is reliable.