Classification of flexible three-fingered hand grasping pattern based on BP neural network

In robotic application, flexible actuator is the end terminal parts. Rigid actuators are accurate but have poor security and practicability. This paper designed a new type of pneumatic dexterous hand - flexible three-fingered hand. The flexible three-fingered hand grasping pattern can be divided into griping, grasping and holding. The pattern classification of flexible three-fingered hand is designed based on the BP neural network. The network training results show that the proposed classification can determine the operation pattern of flexible three-fingered hand, according to the characteristics of the specific operating parameter vector of the target.