Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Towards Automated and Predictive Network-Level Energy Profiling in Reconfigurable IoT Systems

Mohammud J. Bocus, Senhui Qiu, Robert J. Piechocki, Kerstin Eder

TL;DR

This work addresses the challenge of energy efficiency in distributed IoT networks by introducing a measurement-driven, network-level energy profiling framework that unifies BLE and VLC with environmental sensing and E-ink subsystems. It delivers end-to-end data collection, cross-layer energy estimation, and ML-ready datasets that enable AI-based predictive optimization, validated through hardware-in-the-loop experiments in a real testbed. Key contributions include node/gateway/AP energy models with high predictive accuracy (e.g., $R^2$ around 0.99 for network-level energy), a fully automated profiling system, and a demonstration of significant energy savings by reconfiguring operational states and duty cycles. The framework supports proactive energy management and sustainable IoT design by enabling accurate forecasting and optimization across heterogeneous multi-protocol deployments.

Abstract

Energy efficiency has emerged as a defining constraint in the evolution of sustainable Internet of Things (IoT) networks. This work moves beyond simulation-based or device-centric studies to deliver measurement-driven, network-level smart energy analysis. The proposed system enables end-to-end visibility of energy flows across distributed IoT infrastructures, uniting Bluetooth Low Energy (BLE) and Visible Light Communication (VLC) modes with environmental sensing and E-ink display subsystems under a unified profiling and prediction platform. Through automated, time-synchronized instrumentation, the framework captures fine-grained energy dynamics across both node and gateway layers. We developed a suite of tools that generate energy datasets for IoT ecosystems, addressing the scarcity of such data and enabling AI-based predictive and adaptive energy optimization. Validated within a network-level IoT testbed, the approach demonstrates robust performance under real operating conditions.

Towards Automated and Predictive Network-Level Energy Profiling in Reconfigurable IoT Systems

TL;DR

This work addresses the challenge of energy efficiency in distributed IoT networks by introducing a measurement-driven, network-level energy profiling framework that unifies BLE and VLC with environmental sensing and E-ink subsystems. It delivers end-to-end data collection, cross-layer energy estimation, and ML-ready datasets that enable AI-based predictive optimization, validated through hardware-in-the-loop experiments in a real testbed. Key contributions include node/gateway/AP energy models with high predictive accuracy (e.g., around 0.99 for network-level energy), a fully automated profiling system, and a demonstration of significant energy savings by reconfiguring operational states and duty cycles. The framework supports proactive energy management and sustainable IoT design by enabling accurate forecasting and optimization across heterogeneous multi-protocol deployments.

Abstract

Energy efficiency has emerged as a defining constraint in the evolution of sustainable Internet of Things (IoT) networks. This work moves beyond simulation-based or device-centric studies to deliver measurement-driven, network-level smart energy analysis. The proposed system enables end-to-end visibility of energy flows across distributed IoT infrastructures, uniting Bluetooth Low Energy (BLE) and Visible Light Communication (VLC) modes with environmental sensing and E-ink display subsystems under a unified profiling and prediction platform. Through automated, time-synchronized instrumentation, the framework captures fine-grained energy dynamics across both node and gateway layers. We developed a suite of tools that generate energy datasets for IoT ecosystems, addressing the scarcity of such data and enabling AI-based predictive and adaptive energy optimization. Validated within a network-level IoT testbed, the approach demonstrates robust performance under real operating conditions.

Paper Structure

This paper contains 16 sections, 13 equations, 9 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. System Description
  4. SUPERIOT Hardware
  5. Node-Level Energy Measurement, Analysis, Modeling and Prediction
  6. Energy Consumption Overview of RIoT Node
  7. Gateway/Access Point Energy Measurement, Analysis, Modeling and Prediction
  8. BBB AP Energy Consumption
  9. Impact of LED Brightness on Average Current Consumption
  10. Energy Modeling
  11. Network-Level Energy Measurement, Analysis, Modeling and Prediction
  12. Measurement Setup
  13. Dataset Preprocessing
  14. Machine Learning Models for Predicting Network-Level Energy Consumption
  15. Conclusion
  16. ...and 1 more sections

Figures (9)

  • Figure 1: Illustration of (a) Custom-engineered, multi-functional Si-based RIoT node, (b) Mini-lamp gateway supporting dual-mode communication (BLE and VLC), and (c) BBB Access Point (BBB platform + mini-lamp gateway mounted on a Cape).
  • Figure 2: Node-level energy consumption prediction App.
  • Figure 3: Predicted energy consumption of Si-based node over a 24-hour period under various scenarios.
  • Figure 4: Comparison of the average current consumption between BBB AP and mini lamp gateway under three test cases.
  • Figure 5: Current consumption analysis of mini lamp gateway during different: (a) BLE scanning duty cycle values, (b) BLE connection intervals (VLC off in both cases).
  • ...and 4 more figures