Fig. 5: The ability of the multifunction tactile system for material classification and roughness discrimination.

a Photographs of six different objects with various materials and similar roughness, containing PET film, PI film, silicon wafer, stainless steel sheet, glass and silk film. b Details of the corresponding responses of the bioinspired sensor during sliding on above objects with the constant normal force of 2 N and sliding velocity of 100 mm min−1. c Confusion matrix of materials discrimination from the model trained by XGBoost. Predicted label refers to the recognized results, and true label refers to the true materials. d SEM images of six different silicon wafers with various roughness with the same depth (40 μm) and six kinds of spacing and width. Scale bar represents 500 μm. e Details of the corresponding responses of the bioinspired sensor during sliding on above silicon wafers with the constant normal force of 2 N and sliding velocity of 100 mm min−1. f Confusion matrix of roughness discrimination from the model made XGBoost. Predicted label refers to the recognized results, and true label refers to the true roughness. The color bar represents the normalized accuracy.