Tactile Expert System Using a Parallel-Fingered Hand Fitted with Three-Axis Tactile Sensors

Abstract

In an effort to improve the intelligence of a robot fitted with tactile sensors, we have developed a tactile sensor array which detects the distribution of three applied force vector components. Here, we introduce formulas for basic tactile information such as applied force and moment, centroid of distributed pressure, contact pattern, contact state and mechanical properties, obtained through contact of the tactile sensor with an object. We then produce a tactile expert system which is capable of recognizing the material, shape and stability of the object by means of a reasonably combining the basic tactile information. Recognition experiments show that the system can infer the material and shape of seven objects and that it can control the hand in response to the different kinds of grasping instability.