Data-Driven Energy Estimation of Individual Instructions in User-Defined Robot Programs for Collaborative Robots
Juan Heredia, Christian Schlette, Mikkel Baun Kjærgaard
Abstract
Traditionally, the objective of industrial production focuses on fast and low-cost production, regardless of resources and energy consumption (EC). However, in alignment with the UN Sustainable Development Goals (SDG), governments worldwide have proposed regulations to reduce resources and energy. In their production lines, an increasing number of companies are using collaborative robots (cobots). Cobots are programmed to accomplish their task as fast as possible, ignoring the robot's EC. This letter estimates the cobot EC from individual instructions of user-defined robot programs. Thus, the user has an additional design parameter to create energy-optimal programs. In the literature, current EC estimation models for manipulators are not reliable or have not been assessed to test the model's reliability. Our modeling methodology possesses three steps: motion planning, dynamic model, and EC model. Using cobots of different sizes (UR3e and UR10e) and loading, we collected over 55000 samples per case and trained the model to identify the model's unknown parameters. The model estimated the power consumption of a testing dataset with a maximum RMS error of 6 [W] - 3.85%. In the final experiment, the complete system was tested using a user-defined program composed of six instructions. The results showed an accurate estimation of the power profile with an RMS error of 2.39 [W] and 4.23 [W] for UR3e and UR10e.