Tactile Recognition of Both Shapes and Materials with Automatic Feature Optimization-Enabled Meta Learning
Hongliang Zhao, Wenhui Yang, Yang Chen, Zhuorui Wang, Baiheng Liu, Longhui Qin
AI summary
Problem
Tactile perception for robots suffers from severe data scarcity and costly data collection, making traditional deep learning models prone to overfitting and slow adaptation when only few examples are available.
Approach
The authors propose AFOP-ML, a meta-learning framework that automatically ranks and selects the optimal subset of tactile features using Neighborhood Component Analysis and an episodic scan, then feeds them into a lightweight prototypical network for rapid few-shot adaptation.
Key results
- 96.08% accuracy in 5-way-1-shot classification
- 88.74% accuracy in extreme 36-way-1-shot scenario
- Strong generalization across unseen shapes, materials, and perturbations
- Reduced adaptation latency (~391 ms/episode) versus baselines
Why it matters
It provides a data-efficient, rapidly adaptable solution for robotic tactile perception, enabling dexterous manipulation in contact-rich scenarios where visual cues are limited and training data is scarce.
Abstract
Tactile perception is indispensable for robots to implement various manipulations dexterously, especially in contact-rich scenarios. However, alongside the development of deep learning techniques, it meanwhile suffers from training data scarcity and a time-consuming learning process in practical applications since the collection of a large amount of tactile data is costly and sometimes even impossible. Hence, we propose an automatic feature optimization-enabled prototypical network to realize meta-learning, i.e., AFOP-ML framework. As a “learn to learn” network, it not only adapts to new unseen classes rapidly with few-shot, but also learns how to determine the optimal feature space automatically. Based on the four-channel signals acquired from a tactile finger, both shapes and materials are recognized. On a 36-category benchmark, it outperforms several existing approaches by attaining an accuracy of 96.08% in 5-way-1-shot scenario, where only 1 example is available for training. It still remains 88.7% in the extreme 36-way-1-shot case. The generalization ability is further validated through three groups of experiment involving unseen shapes, materials and force/speed perturbations. More insights are additionally provided by this work for the interpretation of recognition tasks and improved design of tactile sensors.