A Tactile Recognition System Mimicking Human Mechanism for Recognizing Surface Roughness

A mathematical model was formulated on the basis of results from psychophysical experiments in which human subjects discriminated fine steps on aluminum plates. The mathematical model emulated the real neuron discharge caused when a membrane potential exceeds a threshold. This membrane potential was determined by spatial and temporal summations of postsynaptic potential. To evaluate the mathematical model for surface texture recognition by robots, we performed a series of surface-detection experiments using a robotic manipulator equipped with an optical three-axis tactile sensor. The single sensor cell of this sensor consisted of a columnar feeler and a 2-by-2 array of conical feelers. The three-axis force was calculated from the area-sum and area-difference of the conical feelers’ contact areas. The robotic manipulator rubbed the tactile sensor on four brass plates with step heights of 0, 0.05, 0.1 and 0.2mm. Results showed that the mathematical model could distinguish these step heights in real time.