Haptic Based Surface Texture Classification Using Machine Learning Techniques

This paper explores the application of machine learning techniques for object classification based on the haptic sensation of surface textures. The haptic sensation of different surfaces was recorded using an actuator as the position and reaction force response variations. The reaction force from the surface was obtained through an estimation using sensorless sensing technique of reaction force observer. Using the gathered data for five different surface classes, feature extraction was done and feature correlation was identified. A Random Forest (RF) classifier model and a Convolutional Neural Network (CNN) model were trained using selected features. Through testing and measuring performances, it was identified that the CNN model outperformed the RF model in terms of accuracy and speed. However, for training the CNN model requires a high amount of computational resources. Despite this drawback, the CNN algorithm is apparently the most suitable machine learning model for this application. This approach has potential applications in industrial processes such as quality assurance and wear and tear assessment.