Related Experiment Video
Updated: Aug 13, 2025
06:37
Author Spotlight: Addressing Technical and Subjective Challenges in Measuring Classroom Attention
Published on: December 15, 2023
4.0K
Large-Scale Tactile Detection System Based on Supervised Learning for Service Robots Human Interaction
Fábio Cunha1,2, Tiago Ribeiro1,2, Gil Lopes3
1Industrial Electronics Department, University of Minho, 4800-058 Guimarães, Portugal.
Sensors (Basel, Switzerland)
|January 21, 2023
Summary
A new tactile detection system uses machine learning and computer vision to map forearm sleeve surfaces for enhanced human-robot interaction. This system enables collaborative robots like CHARMIE to detect contact and perform specific behaviors, improving safety and functionality.
Area of Science:
- Robotics
- Artificial Intelligence
- Human-Computer Interaction
Background:
- Service robots require advanced sensory capabilities for safe and effective human interaction.
- Existing tactile systems often lack comprehensive coverage or adaptability for dynamic environments.
Purpose of the Study:
- To develop a large-scale tactile detection system for a collaborative robot (CHARMIE).
- To enable robust forearm surface contact detection for improved human-robot interaction.
Main Methods:
- Utilized a soft structure with Machine Learning and Computer Vision algorithms.
- Developed an Artificial Neural Network (ANN), the Section Detection Network (SDN), using Supervised Learning.
- Employed Stratified K-Fold Cross Validation (SKFCV) for dataset division and model evaluation.
Main Results:
- Achieved a test data accuracy higher than 80% for contact location prediction.
- Demonstrated real-time evaluation via a graphical interface showing predicted class and probability.
- The system maps forearm sleeve surfaces, enabling segmented contact detection.
Conclusions:
- The proposed tactile detection system significantly enhances human-robot interaction capabilities.
- This technology holds substantial potential for service robots in domestic and medical settings.
- The system's modular design allows for configurable functionalities based on detected contact points.
