Tactile Description

Title: Making sense of a robotic touch

Abstract:

The accurate description of how robots perceive tactile information is not widely known in literature. This paper maps the psychophysical perception of tactile information from humans onto robots. We hypothesize that human sensation of rubbing an object can be characterized by psychophysical evaluation of touch. Based upon this hypothesis, a novel descriptor is proposed that captures the exploratory behaviour of a sensor on a surface. We describe this robotic perception of surfaces along different temporal frequency bands and show that this perception is robust against noise. Using different types of tactile sensors, we show that the proposed description accurately maps the intrinsic properties of a surface into robotic perception. This mapping exists at an abstract level of perception, thereby making the description robust against the subjective constraints of the sensor.

Projection of feature space along first 3 principal components

Data Acquisition using Baxter Robot Comparison between different features for Fabric Classification

Syed Sohaib Ali, Abdul Attayeb Khan, Ahmad Mahmood Tahir and Junaid Imtiaz. "Making sense of a robotic touch" under preparation for Journal of Testing and Evaluation. (PDF and Code coming soon!)