Data-Driven Modeling for Privacy-Preserving Energy Prediction of Industrial Robots

This study presents a data-driven, privacy-preserving framework for predicting the energy consumption of industrial robots. First, we compare different machine-learning based approaches for the modeling problem and show that high accuracy can be achieved. Second, to protect sensitive data in collaborative workflows, privacy-preserving machine learning (ppML) techniques based on multi-party computation (MPC) are applied to the trained models. The approach enables accurate energy modeling while maintaining data confidentiality, which is critical in industrial settings where intellectual property protection is essential, thus promoting safe and efficient energy optimization in collaborative workflows.