Abstract
We have developed an artificial finger which have human tactile sensibility function, such as be desired for some kind of robots. This finger can recognizes some materials pressing and tracing object with the arm actuated by the pneumatic cylinder. There are many factors in recognizing what touched material was. The recognition of materials is run by the neural network system which can be usefully to obscure and complicated control. The sensor in the finger has been used a piezo-electric element and a strain gauge with a thermo-couple respectively before. However only one of these sensors can be installed for one finger and the two or more fingers must be used to recognize one object. Then, the complete recognition is difficult for the reduction of measurement accuracy caused by individual try of many finger and the problem of the setting of output data in the neural network system. Therefore, the purpose in this report is to unify two or more sensors into one finger, and to fulfill the human tactile sensibility function more similar to human than previous subj ect. In this report, the improvement of the sensor and the device of the neural network are shown.
