Data-Driven Energy Modeling of Industrial IoT Systems: A Benchmarking Approach
Dimitris Kallis, Moysis Symeonides, Marios D. Dikaiakos · 2025
Read Data-Driven Energy Modeling of Industrial IoT Systems: A Benchmarking Approach by Dimitris Kallis, Moysis Symeonides, Marios D. Dikaiakos (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-2505-02543.txt.
Introduction
Dimitris Kallis, Moysis Symeonides, and Marios Dikaiakos benchmarked the energy appetite of an industrial internet-of-things system in a 2025 paper. Factories are filling with connected devices, robotic arms, conveyors, smart cameras, that automate making. Their energy bills are becoming a real operating cost that physics-based models predict poorly. The authors put forth a benchmarking methodology for measuring and analyzing device and application power demands. Then show it on a teaching-scale cyber-real system: a robotic arm, a conveyor belt, a smart camera, and a work out node.
Micro-benchmarks and an end-to-end application generate a rich data set of speed and power readings. Used to train machine learning models that predict energy use from application and system traits. Data-driven energy models, the authors argue, are the useful path to greener industrial automation.
Catalog record
| CREATOR | Dimitris Kallis, Moysis Symeonides, Marios D. Dikaiakos |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 36 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2505-02543
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Data-Driven Energy Modeling of Industrial IoT Systems: A Benchmarking Approach Thanks: This is a pre-print of a paper accepted for presentation at the 30th IEEE Symposium on Computers and Communications (ISCC) 2025. Dimitris Kallis, Moysis Symeonides, Marios D. Dikaiakos Abstract The widespread adoption of IoT has driven the development of cyber-physical systems (CPS) in industrial environments, leveraging Industrial IoTs (IIoTs) to automate manufacturing processes and enhance productivity. The transition to autonomous systems introduces significant operational costs, particularly in terms of energy consumption. Accurate modeling and prediction of IIoT energy requirements are critical, but traditional physics- and engineering-based approaches often fall short in addressing these challenges comprehensively. In this paper, we propose a novel methodology for benchmarking and analyzing IIoT devices and applications to uncover insights into their power demands, energy consumption, and performance. To demonstrate this methodology, we develop a comprehensive framework and apply it to study an industrial CPS comprising an educational robotic arm, a conveyor belt, a smart camera, and a compute node. By creating micro-benchmarks and an end-to-end application within this framework, we create an extensive performance and power consumption dataset, which we use to train and analyze ML models for predicting energy usage from features of the application and the CPS system. The proposed methodology and framework provide valuable insights into the energy dynamics of industrial CPS, offering practical implications for researchers and practitioners aiming to enhance the efficiency and sustainability of IIoT-driven automation. 1 Introduction Exponential advances in Internet of Things (IoT) technologies have led to a rapid proliferation of IoT applications deployed in numerous domains, such as transportation, healthcare, industrial control, etc. Modern industrial infrastructures increasingly comprise cyber-physical systems (CPS), which integrate industrial devices with computational components and processes [ 1 ] . Typically, CPS consist of IoT sensors, which monitor physical processes and systems to collect data reflecting their state, and of IoT actuators, which control physical systems to complete selected tasks. The management of CPS systems is performed by software components, which implement various computational processes like data analytics and Machine Learning (ML) inference to provide CPS-component control and coordination. These software components are typically deployed in computing nodes within small data centers located near the IoT installations [ 2 ] . As CPS proliferate in distribution and scale, their operation is expected to take an increasing portion of the total energy consumption of IT systems, contributing significantly to global electricity consumption and CO_{2} emissions. Consequently, many recent research efforts focus on analysing and modeling the energy profile of edge computing devices and application workloads, and on adding energy-efficiency features to edge computing operations [ 3 , 4 , 5 , 6 ] . Most of these efforts, however, focus on profiling computational workloads running on individual IoT computing devices. Few studies focus on analyzing CPS installations that comprise both computing as well as sensing and actuating components working in tandem to perform complex tasks. Such scenarios are nevertheless prevalent in industrial infrastructures and applications, particularly where Industrial IoT (IIoT) deployments automate manufacturing processes within industrial production pipelines. Such automation enables modern factories to operate around the clock, reducing personnel costs and enhancing productivity. With the prices of robotic arms dropping fast (46.2% reduction between 2017 and 2021 [ 7 ] ), analysts predict that such devices will become ubiquitous and could unleash a huge potential for the global economy [ 8 ] . Given that power requirements significantly impact the operational costs of IIoT deployments, it is crucial to explore the power, energy, and performance of such applications, considering both their computational and physical components and performed tasks. However, studies of IIoT deployments often rely on simplistic models or focus on specific industrial machinery that do not consider real-world applications [ 9 , 10 , 11 ] . More comprehensive and realistic models are required to better predict energy consumption, improving the efficiency and cost-effectiveness of IIoT systems. In this paper, we present a methodology and a benchmarking framework for analyzing the energy profile of CPS configurations in industrial automation systems, and for developing predictive models of their energy consumption. We deploy and demonstrate this framework to analyze the power and energy profile of an industrial CPS system, comprising a robotic arm, a conveyor belt, and a smart camera, collaboratively performing automated object sorting based on visual attributes such as color. The methodology is supported by our benchmarking and monitoring framework, which analyzes the energy profile of instructions submitted for execution by the application to the physical (e.g., robotic arm, belt) and digital (e.g., camera) components of the installation. The analysis maps component energy profiles, links them to operational parameters, and trains models to predict the energy profile of end-to-end applications based on features like speed, acceleration, and weight. The rest of the paper is structured as follows: Sec. 2 introduces the related work and Sec. 3 our cyber-physical setup. Next, Sec. 4 and 5 show our methodology and its workloads, respectively. In Sec. 6 and 7 , we apply exploratory analysis and the AI/ML training, while Sec. 8 concludes the paper. 2 Related Work The initial efforts to model the energy consumption of mechanical robotic components are related to the optimization of multi-robot systems [ 9 ] . However, the models and the analysis of these cases were relatively simple, taking into account only the initial and final trajectory points of a movement without considering the user-defined application. Recognizing this gap, Heredia et al [ 12 ] introduced an instruction-based approach to model the energy consumption of robots. By determining the energy consumed by each instruction, this approach enables the calculation of the overall energy consumption for any user-defined program. Building on this foundation, Heredia et al in [ 13 ] , proposed a novel optimization strategy tailored for lightweight robots, incorporating three energy-minimizing techniques: manufacturer command optimization, motion time determination optimization, and dissipative energy reduction. Additionally, in [ 14 ] , they developed an energy consumption disaggregation pipeline, demonstrating its application on four robotic arms from various manufacturers. Beyond mathematical formulation and modeling of robotic components’ energy consumption, other approaches have emerged. For instance, the authors of [ 15 ] propose a methodology for developing models that map parameters such as joint torque and joint angular velocity of a robotic welding machine subsystem to its corresponding electric current and voltage. The authors in [ 16 ] , introduce a digital twin system to model the energy characteristics of robotic components, enabling the replication and analysis of results within a simulated environment. Yan et al [ 17 ] and Ming et al [ 18 ] , adopt deep learning (DL) and transfer learning techniques to to model the energy profiles of robotic systems, with Yan et al. focusing on fine-tuning a neural network trained on one robotic arm to achieve high accuracy on a different system, while Ming et al. employ a ResNet-based model to evaluate the effectiveness of transfer learning in energy prediction tasks. In summary, mathematical formulations cannot model complex robotic systems that involve multiple parameters and cyber-physical components. Despite offering improved results in specific scenarios, approaches like digital twins, simulations, and DL, suffer from critical limitations: (i) they do not offer any methodology generalization; (ii) they are