A hybrid mechanism modeling and data-driven method for energy prediction and optimization for a class of industrial robots

Abstract

Predicting and optimizing the energy consumption of industrial robots (IRs) helps reduce their operating costs and achieve environmental protection. However, existing methods are often constrained by IR controller inputs or rely heavily on measured data, limiting their applicability to ordinary production-line IRs. To overcome these challenges, this paper proposes a hybrid robot energy model (HREM) for a class of commonly used elbow-type IRs, which integrates a robot simplified mechanism-based model (RSM) with a neural network for residual energy compensation. This approach combines the strengths of mechanism modeling and data-driven approaches. Unlike purely data-driven methods, HREM reduces dependence on training data and enables accurate prediction and optimization across untrained joint positions. In the mechanism modeling part, RSM parameters can be identified using only power-supply-side data, and an optimization algorithm combining gradient descent and genetic algorithms (GD-GA) is introduced to improve identification efficiency. Experimental results demonstrate that HREM achieves higher prediction accuracy and greater energy savings compared with data-driven methods, making it a practical solution for large-scale industrial deployment.