Recognition of surface texture with wearable tactile sensor array: A pilot Study

Discrimination of surface textures and their surface roughness using tactile sensors have attracted increasing attention. Highly sensitive tactile sensors with the ability to recognize and discriminate the surface textures and roughness of grasped objects are crucial for intelligent robotics. This paper presents a methodology by using the developed WMB model (W-M function and Beam-Bundle Theory) and an algorithm based on artificial neural network to study the performance of a flexible tactile sensor for surface texture classification. For the WMB model, the quasi-3D surfaces of specific objects are reconstructed based on W-M function and basic statistical theory. A simplified Beam-Bundle Model is utilized to represent the cover layer of the sensor and simulates the normal force fluctuations during sliding movements. According to the simulation results, surface textures can be classified by the characteristic frequency cluster (CFC) existing in the fluctuation of curve’s spectrum. As an experiment, an artificial neural network is established to classify surface textures based on voltage signals from the tactile sensor. An MAF array represents the CFC information and improves the classification accuracy from 78 % to 82 %. The results demonstrate the effectiveness of the proposed WMB model and that it provides a new method of analysis involving robotic tactile interactions.