Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

On Using Neural Networks to Learn Safety Speed Reduction in Human-Robot Collaboration: A Comparative Analysis

Marco Faroni, Alessio Spanò, Andrea M. Zanchettin, Paolo Rocco

TL;DR

This work tackles the unpredictable slowdown induced by safety mechanisms in human-robot collaboration by learning the safety scaling function directly from process data. It shows that a simple feed-forward network can accurately predict the robot's slowdown, enabling better cycle-time estimation and more effective scheduling in collaborative tasks. Through simulation and a real-world pick-and-pack use case, classification-based one-step predictions outperform regression, and horizon-averaged scaling can be learned for longer horizons. The results suggest a practical path to data-driven safety-aware planning in HRC without relying on rigid predefined safety models.

Abstract

In Human-Robot Collaboration, safety mechanisms such as Speed and Separation Monitoring and Power and Force Limitation dynamically adjust the robot's speed based on human proximity. While essential for risk reduction, these mechanisms introduce slowdowns that makes cycle time estimation a hard task and impact job scheduling efficiency. Existing methods for estimating cycle times or designing schedulers often rely on predefined safety models, which may not accurately reflect real-world safety implementations, as these depend on case-specific risk assessments. In this paper, we propose a deep learning approach to predict the robot's safety scaling factor directly from process execution data. We analyze multiple neural network architectures and demonstrate that a simple feed-forward network effectively estimates the robot's slowdown. This capability is crucial for improving cycle time predictions and designing more effective scheduling algorithms in collaborative robotic environments.

On Using Neural Networks to Learn Safety Speed Reduction in Human-Robot Collaboration: A Comparative Analysis

TL;DR

This work tackles the unpredictable slowdown induced by safety mechanisms in human-robot collaboration by learning the safety scaling function directly from process data. It shows that a simple feed-forward network can accurately predict the robot's slowdown, enabling better cycle-time estimation and more effective scheduling in collaborative tasks. Through simulation and a real-world pick-and-pack use case, classification-based one-step predictions outperform regression, and horizon-averaged scaling can be learned for longer horizons. The results suggest a practical path to data-driven safety-aware planning in HRC without relying on rigid predefined safety models.

Abstract

In Human-Robot Collaboration, safety mechanisms such as Speed and Separation Monitoring and Power and Force Limitation dynamically adjust the robot's speed based on human proximity. While essential for risk reduction, these mechanisms introduce slowdowns that makes cycle time estimation a hard task and impact job scheduling efficiency. Existing methods for estimating cycle times or designing schedulers often rely on predefined safety models, which may not accurately reflect real-world safety implementations, as these depend on case-specific risk assessments. In this paper, we propose a deep learning approach to predict the robot's safety scaling factor directly from process execution data. We analyze multiple neural network architectures and demonstrate that a simple feed-forward network effectively estimates the robot's slowdown. This capability is crucial for improving cycle time predictions and designing more effective scheduling algorithms in collaborative robotic environments.

Paper Structure

This paper contains 18 sections, 14 equations, 8 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related work
  3. Problem Statement
  4. Methods
  5. Data Collection and Self-Labeling via Clustering
  6. One-step scaling prediction
  7. Classification network
  8. Regression
  9. N-step scaling prediction
  10. Average scaling prediction
  11. Experiments
  12. Data collection
  13. Results
  14. One-step scaling prediction
  15. N-step scaling prediction
  16. ...and 3 more sections

Figures (8)

  • Figure 1: Neural network architecture.
  • Figure 2: Safety speed scaling function. Simulations: $d_1$=1.2 m, $d_2$=1.5 m, $d_3$=1.9 m, $d_4$=2.1 m. Real-world: $d_1$=0.6 m, $d_2$=0.8 m, $d_3$=1.2 m, $d_4$=1.6 m.
  • Figure 3: Simulation scenario. Overview of the working space (left). Robot pick&place example (middle). Human picking and placing nominal positions (right).
  • Figure 4: Accuracy of the classification network.
  • Figure 5: LSTM-based networks for $N$-step predictions.
  • ...and 3 more figures