Energy-Sustainable Fog System for Mobile Web Services in Infrastructure-Less Environments

Qiushi Zheng, Jiong Jin, Tiehua Zhang, Jianhua Li, Longxiang Gao, Yong Xiang

Introduction

In urban environments, ubiquitous cloud connectivity is often taken for granted. However, in remote regions such as Australia’s national parks, the absence of cellular coverage and electrical grids creates a significant "infrastructure-less" gap. This lack of connectivity poses risks for tourists requiring navigation or emergency services and hinders park administrators' ability to monitor environmental conditions and wildlife. This paper proposes a self-powered, hierarchical fog computing system designed to bridge this gap by providing high-performance web services and local edge intelligence without relying on traditional infrastructure.

System Architecture

The proposed solution utilizes a four-layer hierarchical model to manage data and services efficiently across remote distances. This architecture consists of the Cloud (for long-term backups), Management Nodes (for coordination), Fog Nodes (for local service delivery), and IoT/End-user devices.

Hierarchical architecture of the fog computing system, showing the relationship between cloud, management nodes, and energy-sustainable fog nodes.

By processing data at the "edge"—within 10ms of the user—rather than sending it to a distant cloud server with 100ms+ latency, the system ensures a responsive experience for interactive web services. Furthermore, a non-linear bandwidth allocation strategy prioritizes emergency communications, ensuring that high-traffic usage by tourists does not impede critical SOS messages.

Hardware and Energy Sustainability

Each fog node is built around a Raspberry Pi 3B, integrated with various communication modules including Nanobeam (a 5GHz long-range Wi-Fi bridge for links over 5km), ZigBee for low-power sensors, and standard Wi-Fi for user access.

Detailed hardware diagram of a Fog Node, illustrating the integration of the Raspberry Pi with communication modules and the solar power system.

A critical challenge for off-grid nodes is maintaining a steady power supply. The system is powered by a Photovoltaic (PV) system with AGM batteries. To ensure year-round reliability, the authors optimized the solar panel tilt for the winter solstice—the period of lowest solar yield. In Melbourne, this resulted in an optimal tilt of 65°, significantly steeper than the 37° typically used for grid-tied systems, preventing system failure during the darkest months.

Edge Intelligence and Animal Detection

To minimize data transmission, the nodes perform local image processing using a lightweight Convolutional Neural Network (LeNet-5). This allows the system to detect and verify animal species (such as foxes, cats, and wallabies) directly at the node. Using a dataset of 1,000 photographs from Parks Victoria, the model achieved 95% accuracy in animal verification. While effective, the local processing causes significant CPU usage (70%) and thermal spikes, highlighting a need for robust thermal management in the node's 3D-printed housing.

Key Results and Performance

Field tests and simulations demonstrated the system's superiority over traditional remote connectivity options:

  • Latency: Offloading calculations to fog nodes dramatically reduced application loop latency compared to cloud-only strategies.
  • Throughput: In field tests at Wattle Park, Melbourne, a single node provided 7.7 Mbps of bandwidth. Even with 20 concurrent users, it maintained 1.9 Mbps—sufficient for 720p video and significantly faster than typical rural 3G speeds (<1 Mbps).
  • Energy Efficiency: The fog-centric approach consumed significantly less energy per task (~0.8 Megajoules for cloud vs. a small fraction for fog nodes).

Conclusion and Future Work

This energy-sustainable fog system demonstrates that high-performance web and AI services can be deployed in the most remote environments. Future research will focus on distributed CNNs to share computational loads across multiple nodes and the development of self-healing routing protocols to scale the system into massive mesh networks covering entire national parks.

Summary updated

Abstract

Nowadays, critical information services such as emergency management and in-situation navigation rely heavily on the assumption of a reliable networking infrastructure and stable cloud processing, which is ineffective in infrastructure-less environments where disrupted or even no telecommunication connectivity is the norm. On the other hand, fog computing is an extension of cloud computing that is at the physical proximity of end-users to enable local storage, computing, and various forms of communication between devices and users. To this end, we design a fog system spanning hardware, software and networking, thus meeting the requirements of various stakeholders in difficult surroundings. As a complement to a cloud-centred solution, this system is geo-distributed, self-powered, self-managed, location-aware, highly efficient and able to provide situational information services without Internet connectivity. The proposed system has been implemented in national parklands of Australia to achieve a personalized information service, emergency management and in-park navigation for all types of parkland users.
BibTeX
@article{zheng2019energy,
  author = {Qiushi Zheng and Jiong Jin and Tiehua Zhang and Jianhua Li and Longxiang Gao and Yong Xiang},
  title = {{Energy-Sustainable Fog System for Mobile Web Services in Infrastructure-Less Environments}},
  journal = {IEEE Access},
  year = {2019},
  doi = {10.1109/access.2019.2951162},
}
Figure