Researchers have developed an Energy-aware Adaptive Virtualization and Migration (EAVM) protocol that combines federated deep reinforcement learning (FDRL) with hybrid solar-radio frequency (RF) energy harvesting, targeting sustainability challenges in wireless sensor networks (WSNs).
Using large-scale simulations, the team shows that their approach can lower overall energy consumption by about 17 % and extend network lifetime by more than 20 % compared with existing methods, while also improving response time and resource utilization.
Get all the details: Grab your PDF here!
Wireless sensor networks underpin many IoT applications, from smart buildings to intelligent transportation systems. But most sensor nodes operate under severe energy constraints, often relying on batteries that limit system lifetime and scalability.
Green IoT research aims to address this by reducing power consumption across the network while maintaining service quality.
Two techniques are central to this effort: virtualization, which abstracts physical resources into flexible virtual entities, and migration, which dynamically reallocates workloads among nodes to balance energy and performance demands.
Managing these processes efficiently becomes increasingly difficult as networks scale to hundreds or thousands of nodes.
A Learning-Based Approach to Resource Management
The proposed EAVM protocol tackles this problem by using federated deep reinforcement learning, allowing sensor nodes to learn energy-aware resource allocation policies locally while sharing only model updates rather than raw data. This decentralized design avoids the overhead and fragility of centralized control.
To further reduce reliance on batteries, the framework incorporates hybrid solar–RF energy harvesting, modeled within the simulations to reflect intermittent and variable energy availability.
By jointly considering energy supply, workload demand, and user connectivity, EAVM enables predictive migration and adaptive virtualization decisions as changes occur.
The protocol is organized as a multi-layer architecture, spanning energy-aware sensing at the physical layer, intelligent resource allocation at the management layer, and reliable service delivery at the application layer.
Tested at Scale in Simulation
The researchers evaluated EAVM using Contiki Cooja simulations, modeling dense IoT environments with varying node densities and user demands. Across these scenarios, the protocol consistently outperformed both conventional and learning-based comparison methods.
In addition to reduced energy consumption and longer network lifetime, EAVM achieved up to 10 % faster response times, approximately 4 % higher resource utilization, and lower migration overhead, indicating a more efficient use of limited network resources.
Importantly, these gains were observed without centralized coordination, highlighting the scalability of the federated learning approach.
Limitations in a Greener IoT Future
The authors stress that the results are based on simulation rather than real-world experiments. Practical systems may face additional challenges, including hardware heterogeneity, communication noise, and stricter privacy requirements when exchanging learning parameters.
While the study discusses potential applicability to fog-cloud-assisted environments, no explicit fog-cloud implementation was evaluated.
Future work will focus on privacy-preserving federated optimization, carbon-aware scheduling, and context-driven energy forecasting, aiming to bridge the gap between simulation and operational IoT systems.
Journal Reference
Liu, Y., Li, Y. and Ge, N. (2025) Energy-Aware adaptive virtualization and migration protocol for green IoT wireless sensor networks. Scientific Reports, 15, 44944. DOI: 10.1038/s41598-025-28783-z
Disclaimer: The views expressed here are those of the author expressed in their private capacity and do not necessarily represent the views of AZoM.com Limited T/A AZoNetwork the owner and operator of this website. This disclaimer forms part of the Terms and conditions of use of this website.