A Spiking Neural Network Model for Object Recognition with Tactile Properties

Tactile properties such as hardness, roughness, and texture are critical information for robots interacting with their surrounding environment, forming the foundation for various tasks including grasping, autonomous navigation, and collaboration. Inspired by human perception mechanisms, this paper p...

Full description

Saved in:
Bibliographic Details
Published inInternational Conference on Automation, Control and Robotics Engineering (Online) pp. 210 - 215
Main Authors Wang, Fei, Wu, Juan, Pan, Yanxin
Format Conference Proceeding
LanguageEnglish
Published IEEE 16.07.2025
Subjects
Online AccessGet full text
ISSN2997-6278
DOI10.1109/CACRE66141.2025.11119548

Cover

More Information
Summary:Tactile properties such as hardness, roughness, and texture are critical information for robots interacting with their surrounding environment, forming the foundation for various tasks including grasping, autonomous navigation, and collaboration. Inspired by human perception mechanisms, this paper presents a tactile property recognition method based on spiking neural network (SNN). We designed a distributed three-dimensional tactile sensor that mimics the structure of human skin, aiming to collect tactile signals through active exploration modes of pressing and sliding over artificially synthesized samples with varying textures and roughness. The collected signals are converted into pulse sequences using Poisson encoding, with object classification determined by the frequency of spike emissions. Experimental results indicate that the proposed method exhibits a straightforward structure and robust predictive performance, achieving an object classification accuracy of 95.56%, effectively simulating human tactile perception capabilities.
ISSN:2997-6278
DOI:10.1109/CACRE66141.2025.11119548