Research On Perception And Control Technology Of Tendon-driven Space Multi-fingered Dexterous Hand

With the development progress of China's space laboratory projects,in order to enhance the technology of on-orbit service,space intelligent robots are received much attention.Designing space multi-finger dexterous hand of high capability has become a boom in the domain of astronautics.This paper mainly studies the key issues related to the perception and control of tendon-driven multi-fingered dexterous hand.In this paper,it first designs the anthropomorphic structure of palm,thumb,the other four fingers and the position of the thumb.In the sensor system,modeling simulation and mechanism design on tendon tension sensor is mainly made,the simulation result verifies correctness of measurement principle.The structure of six axis force sensor is designed,and the statics,modality and transient characteristics related are also analyzed.As the foundation of finger force control,kinematics and dynamics problems are described in detail.In this part,suitable coordinate system are established on the finger joints.Then using D-H parameter method,the kinematic model of index finger is presented.Taking advantage of the relationship between joint velocity and fingertip velocity derives the Jacobi matrix in differential kinematics.Based on the geometric relation of index finger coupling action between connecting rods,the dynamic equation of the index finger is deduced by Newton-Euler iterative algorithm.Two impedance control approaches are proposed including tendon-space and joint-space schemes for finger compliant control.Joint simulations are performed,which verifies the effectiveness of two impedance control method.The influence of envelope angle,friction and flexibility of the tendon on transmission characteristics of tendon-driven multi-fingered dexterous hand are analyzed.The influence of transmission characteristics of the tendon on finger's force compliance control system is simulated,which verifies joint-space scheme is better than tendon-space scheme when taking into account of friction.At last,the contact constraints and grasping statics are introduced.The concrete design of two algorithms including sequence quadratic programming(SQP)and linear constraint gradient flows are proposed for current designed multi-fingered grasping force optimization.The linear constraint gradient flows algorithm is improved on tendon tension constraint and real-time improvement.A specific planning example is also realized by using SQP and then is numerical verified.