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

Juan Heredia, Christian Schlette, Mikkel Baun Kjærgaard · 2025

Read EcBot: Data-Driven Energy Consumption Open-Source MATLAB Library for Manipulators by Juan Heredia, Christian Schlette, Mikkel Baun Kjærgaard (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2508-06276.txt.

Introduction

Juan Heredia, Christian Schlette, and Mikkel Baun Kjærgaard release EcBot, an open MATLAB library that learns how much power a robot arm uses, in a 2025 paper. Older power models were mostly checked on heavy industrial robots and often miss the mark. EcBot builds its model from real running data instead.

The user supplies the arm's geometry, link masses, and centers of mass, plus logged joint motion and power traces. The library then fits a data-driven model automatically. Tests on four light robots from three makers gave errors of only a few watts on both training and fresh data. The authors offer the tool for energy-aware planning and greener automation.

Catalog record

CREATORJuan Heredia, Christian Schlette, Mikkel Baun Kjærgaard
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE38 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2508-06276

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

EcBot: Data-Driven Energy Consumption Open-Source MATLAB Library for Manipulators Thanks: Juan Heredia, Christian Schlette, Mikkel Baun Kjærgaard Abstract 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 Energy Consumption (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. I Introduction In modern society, the industrial sector has experienced a paradigm shift, moving beyond a sole focus on generating high revenues to promoting sustainable production practices [ 1 , 2 ] . This shift coincides with the growing attention given to robots in various industries, Industrial Robots (IRs) , Collaborative Robots (Cobots) or Leightweight Robots (LWRs) . The integration of robots into industrial settings offers the potential to enhance production rates while concurrently reducing EC through environmentally conscious operations and energy-efficient programming [ 3 ] . While industrial robots have been employed in the industry for a considerable period, with the first robot installed nearly 60 years ago [ 4 ] , a new generation of manipulators has emerged, bringing about a notable market transformation. The distinguishing feature of these new robots is their ability to operate alongside humans in the same environment. To enable this collaboration, these robots are specifically designed with lighter materials and reduced stiffness, allowing them to absorb energy during collisions and minimize potential damage [ 5 ] . Additionally, they incorporate safety protocols aimed at reducing hazards in shared workspaces. These robots are commonly referred to as collaborative robots or lightweight robots. It is important to note that the terms "collaborative robots" [ 6 ] and "lightweight robots" [ 7 ] are often used interchangeably in the existing literature. Figure 1: EcBot MATLAB Library for Dynamic and Energy Consumption Modeling for Manipulators. The focus of this study is to address the significant challenges prevalent in the field of EC modeling for cobots. The analysis begins with a review of the existing academic literature, which predominantly focuses on IRs , highlighting a significant research gap concerning cobots. This EC modeling gap is particularly concerning considering the findings of Gadaleta et al. (2020) [ 8 ] , who argue that current EC models for IRs lack reliability. Furthermore, they report that errors in these models are often concealed in academic literature, exacerbating the problem. The complexity of modern robotics, particularly cobots, significantly contributes to these errors. Verstraten et al. (2016) [ 9 ] underscore this issue, discussing accuracy problems in EC models even for a single DC motor. As the focus shifts to manipulators, these accuracy issues become exponentially more problematic due to the increased complexity of dynamic models, making it challenging to estimate EC accurately. To address these challenges, we previously introduced a modeling approach for estimating the EC of cobots in our prior work [ 10 ] . However, this methodology was only validated on two Universal Robots, specifically the UR10e and UR3e, which share similar mechanical designs. Moreover, the model’s creation involved a semi-manual modeling and training process. When dealing with a robot of different design or structure, the entire process had to be repeated. The novelty of this article lies in the introduction of an automated library that generates, trains, and tests EC models for a wide range of manipulators. This library necessitates specific data inputs, including Denavit-Hartenberg parameters, the mass and center of gravity of the robot’s links, and operational data (refer to Fig. 1 ). This software solution is accessible online at [ 11 ] , where EC models for robots of varying degrees of freedom (DoF) and geometries can be trained and assessed. Additionally, in this paper, we present four case studies where the library is implemented for four distinct robot designs. This highlights the library’s versatility and its applicability to various robot architectures and specifications. The structure of the paper is as follows: Section II provides an overview of the state-of-the-art models from the literature. Section III presents our model in detail. In Section IV, we present case studies showcasing the EC model results for five robots from three different manufacturers: UR3e, UR5e, UR10e, FR3, and Gen3. Finally, Section V presents the conclusions of the article. II Related Work In the literature, there have been several attempts to model the EC of mechatronic systems, ranging from simple 1 DoF mechanisms [ 12 , 9 ] to more complex manipulators such as IRs and Cobots . In this analysis, we will examine the principal attempts from the literature to model the EC of these systems: Figure 2: EcBot: Energy Consumption Data-Driven Modeling Library • Authors in [ 13 ] present an EC model for the ABB IRB 1200 IR . This model focuses on identifying inertia and friction parameters based on simulated data. However, its limitation lies in relying exclusively on simulated data from the manufacturer’s simulation tool, "RoboStudio," for training the model. Additionally, the model only considers the EC of the robot’s motors, neglecting other components. • In a study by [ 14 ] , an EC model for the Motoman MH5L IR is developed using the "Catia Systems Engineering" simulation tool. This model provides a mechanical representation of the robot’s structure and incorporates both mechanical and electrical losses of each motor. However, the testing conducted for this model was limited to short experiments lasting up to 5 seconds. Longer tests are necessary for comprehensive validation. • Another EC model for the KR200 IR is formulated in a study by [ 15 ] . This model encompasses the EC of various components, including the manipulator elements, drive system, and auxiliary components like the computer and cooling system. However, the paper lacks an accuracy analysis as the model has not been subjected to any tests. • A specific EC model for manipulators, using the ABB IRB140 as a case study is presented by [ 16 ] . The methodology involves a three-step process: kinematics, dynamics, and EC calculation. However, this model only considers the EC of the motors, excluding other robot components. Moreover, the validity of the model was not confirmed with real data. • Eggers et al. [ 17 ] present an EC model that incorporates the temperature of the motors. The study uses the KR16 IR to investigate how the robot’s operational temperature influences its EC . The results demonstrate that the EC reduces as the oil temperature of the gearbox increases due to reduced viscosity and mechanical losses. The presented model increases accuracy by accounting for motor temperature. • In a study by [ 18 ] , an EC modeling methodology for manipulators is presented and tested

Read the complete text