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

2025-12-19

AI Summary

In human–robot collaboration, safety-constrained speed limiting leads to inaccurate cycle time prediction and degraded scheduling efficiency. To address this, this paper proposes a runtime data-driven method for learning dynamic safety scaling factors. Unlike conventional approaches relying on predefined safety models, our work is the first to introduce deep learning into this task, enabling end-to-end prediction of safety scaling factors directly from real-world execution data. Through comparative evaluation of multiple neural architectures, we find that a lightweight feedforward neural network achieves high-accuracy modeling of robot real-time deceleration behavior. Experimental results demonstrate that the proposed method significantly improves cycle time prediction accuracy—reducing prediction error by over 40%—thereby enabling superior real-time scheduling decisions. Moreover, it eliminates dependence on manual risk assessment and fixed safety models, enhancing adaptability and operational robustness in dynamic collaborative environments.

Technology Category

• Planning, Routing, and Scheduling: Learning for Planning and Scheduling
• Intelligent Robots: Learning & Optimization for ROB
• Natural Language Processing: Safety and Robustness

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Security and Privacy: Large-scale security measurements

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.

Problem

• Predict robot safety speed scaling from process data
• Improve cycle time estimation in human-robot collaboration
• Enhance scheduling efficiency using accurate slowdown predictions

Innovation

• Deep learning predicts robot safety scaling factor
• Feed-forward network estimates slowdown from execution data
• Improves cycle time predictions for collaborative scheduling