Investigation of a biomimetic fingertip's ability to discriminate fabrics based on surface textures

Tactile sensing is an important ability for a humanoid robot which interacts in an unstructured environment. Such a system needs to sense and evaluate surface properties of objects that it interacts with. Among those properties, surface texture identification is a compulsory ability for certain kind of robot systems such as service robots, medical robots and exploratory robots. Therefore, the tactile system of above type of robots should have the ability to identify and discriminate textures with acceptable accuracy. A biomimetic fingertip that can be used in above kinds of robot tactile systems is introduced. The fingertip has the ability to detect force and vibration modalities. This paper reports the ability of the fingertip system to discriminate multiple materials (six fabrics and aluminium plate) by comparing the differences in their surface texture. The materials were classified using features: variance and power of the accelerometer signal. Moreover an Artificial Neural Network (ANN) classifier was evaluated by using the first 300 Fourier coefficients of the accelerometer signal as features. Above two methods used the raw signals of the accelerometers nearest to the contact area. Finally, an input signal was computed by calculating the covariance of two adjacent accelerometers. This new signal was used to calculate features for the ANN classifier. The results showed that the use of a convoluted signal improved the success rate of discriminating the seven textures.