日本フィジカルAI新聞

世界のフィジカルAIを、日本語で。

週刊ニュースレター購読
arXiv:2508.06276

EcBot: Data-Driven Energy Consumption Open-Source MATLAB Library for Manipulators

EcBot: Data-Driven Energy Consumption Open-Source MATLAB Library for Manipulators

シェア:XThreadsFacebookLINEはてブBluesky

著者: Juan Heredia, Christian Schlette, Mikkel Baun Kjærgaard

分類: cs.RO

原文アブストラクト

Existing literature proposes models for estimating the electrical power of manipulators, yet two primary limitations prevail. First, most models are predominantly tested using traditional industrial robots. Second, these models often lack accuracy. To address these issues, we introduce an open source Matlab-based library designed to automatically generate \ac{ec} models for manipulators. The necessary inputs for the library are Denavit-Hartenberg parameters, link masses, and centers of mass. Additionally, our model is data-driven and requires real operational data, including joint positions, velocities, accelerations, electrical power, and corresponding timestamps. We validated our methodology by testing on four lightweight robots sourced from three distinct manufacturers: Universal Robots, Franka Emika, and Kinova. The model underwent testing, and the results demonstrated an RMSE ranging from 1.42 W to 2.80 W for the training dataset and from 1.45 W to 5.25 W for the testing dataset.