Investigating Active Sampling for Hardness Classification with Vision-Based Tactile Sensors

TL;DR

This paper explores active sampling strategies using vision-based tactile sensors to efficiently classify object hardness, showing significant improvements over random sampling and human performance.

Contribution

It introduces and evaluates uncertainty-driven active sampling methods for hardness classification with tactile sensors, demonstrating their superiority over baseline approaches.

Findings

Active sampling outperforms random sampling in accuracy and stability.

Best method achieves 88.78% accuracy compared to 48% by humans.

Uncertainty-based strategies are effective for tactile hardness classification.

Abstract

One of the most important object properties that humans and robots perceive through touch is hardness. This paper investigates information-theoretic active sampling strategies for sample-efficient hardness classification with vision-based tactile sensors. We evaluate three probabilistic classifier models and two model-uncertainty-based sampling strategies on a robotic setup as well as on a previously published dataset of samples collected by human testers. Our findings indicate that the active sampling approaches, driven by uncertainty metrics, surpass a random sampling baseline in terms of accuracy and stability. Additionally, while in our human study, the participants achieve an average accuracy of 48.00%, our best approach achieves an average accuracy of 88.78% on the same set of objects, demonstrating the effectiveness of vision-based tactile sensors for object hardness classification.