Research On The Structure And Control System Of Humanoid Robot Hand

Humanoid robot hand birthed with prosthetic and developed several generations, and which has entered into a fairly mature development stage. It is a mechatronics system with the feature of high integration and intelligence. As the end-effector of humanoid robot, humanoid robot hand can be designed with powerful structure and rapid movement. As respects of executing task extensively and complexly, due to the technology limitations, such as actuator, sensors and control strategy, the operation capacity of humanoid robot hand cannot be equal to human hand. Thus, besides appearance, structure and dexterous manipulation capacity of human hand should be imitated, advantages of humanoid robot hand needed to combine. Furthermore, control system for humanoid robot hand is urgent to be researched against completing dexterous manipulation task successfully. Researches have been concluded as following aspects in this paper.Modularized structures of humanoid robot hand were designed. Kinematics and statics of finger and the workspace of humanoid robot hand have been analyzed. Furthermore, simulations were carried out, the results shows that humanoid robot hand possesses good mechanical structure properties and large grasping workspace, and which can satisfy the design target.The overall control scheme of humanoid robot hand was designed. Analysis and type selection of main hardware have been developed, such as DSP chip, CPLD chip, power chip and sensor chip. Furthermore, the hardware circuits of main subsystem have been designed, such as power module, motor driven module, communication module and sensor signal processing module.RBF neural network has been improved by combining rival-penalized competitive learning(RPCL) algorithm and recursive orthogonal least-squares(ROLS) algorithm. The neural network dynamic property model of humanoid robot hand has been established and the controller was designed. According to the stability theory analysis and simulation, the control method is stable and the adaptive neural network controller has a very small tracking error. Furthermore, it can control the unknown non-linear dynamic system and external disturbance effectively.The general and main module frameworks of control system solftware were designed. The experiment platform of humanoid robot hand was built. Then, single finger and grasping control experiments were carried out. The results show that the humanoid robot hand possesses good control properties of single finger and good function of grasping some specified objects.