Development of a neuron-inspired tactile information processing model

Abstract

Tactile information processing is essential to robotic manipulation such as grasping, lifting and sliding. We developed a tactile signal processing model by mimicking the firing activity of sensory afferents including slowly adapting (SA) and fast adapting (FA) afferents. This model was validated in publically available data by the classification of the type of nine objects from tactile sensor pressure signals while a robot grasped and releases each object. The classification performance of the developed model was compared with traditional methods based on raw sensor signals, spectral analysis, or statistical analysis. Our model performed better than other models when only one of the tactile sensor data were used and similar to the best model when all the tactile sensor data were used. The proposed model could provide an alternative means to process tactile information in a robotic hand.