Design and Experiment of the NAO Humanoid Robot's Plantar Tactile Sensor for Surface Classification

Abstract

In order to meet the requirement of walking pattern adjustment according to the surface type, aiming at the NAO humanoid robot in this paper, a high resolution tactile sensor based on the force sensing resistor array is designed and fabricated, which can be adhered to the foot of humanoid robot. In the process of robot walking, a series of tactile images are eventually generated by processing the output signal of the tactile sensor. In addition, combined with the k-Nearest Neighbor (kNN) algorithm, a related algorithm for extracting tactile image features is designed, realizing the classification of four typical surfaces. Experiments show that when the number of robot walking steps is greater than 4, the classification accuracy can reach more than 95%.