Energy Aware Routing with Computational Offloading for Wireless Sensor Networks

Adam Barker, Martin SwanyarXiv 2020 · arXiv

Abstract

Wireless Sensor Networks (WSNs) are fundamental to the Internet of Things (IoT) but face significant constraints in battery life and bandwidth. This paper presents EAQCO (Energy Aware Q-routing with Computational Offloading), a novel routing protocol that treats local data processing as a routing decision. By utilizing reinforcement learning and an "energy factor," EAQCO dynamically balances the time and energy costs of local computation against data transmission. Experimental results from a hardware testbed demonstrate that EAQCO reduces delivery latency by approximately 40% and energy consumption by 40% in high-traffic scenarios compared to traditional static routing.

Introduction

As IoT applications generate increasing volumes of sensor data, wireless transmission has become a primary bottleneck in WSNs. Traditional routing protocols typically focus on finding the shortest path to a destination, often overlooking the potential for local processing to reduce data size before transmission. This paper addresses this gap by introducing computational offloading into the routing decision process, allowing nodes to decide whether to forward raw data or process it locally to optimize network performance and longevity.

The EAQCO Protocol

The proposed EAQCO protocol extends the standard Q-routing algorithm by incorporating local computation as a first-class action in the routing table.

Q-learning and Action Selection

In EAQCO, each node maintains a Q-table that estimates the time required to deliver data to a destination via its neighbors. A key innovation is the inclusion of a "Compute Locally" action, which allows a node to perform data reduction in-place. The algorithm for this process is detailed below:

Algorithm 1: Q-routing with computational offloading

By treating the local CPU as a virtual neighbor, the reinforcement learning framework naturally optimizes the computation-vs-communication tradeoff. The rewards for these actions are based on estimated delivery times, as illustrated in the following table:

Action and Reward mapping for the EAQCO routing table

Energy Awareness

To preserve network life, EAQCO introduces an Energy Factor ($ef$). As a node’s battery level decreases, the cost of forwarding increases exponentially. This mechanism automatically shifts the network's priority toward local computation or alternative, lower-hop routes as nodes approach depletion, ensuring a more resilient network topology.

Experimental Evaluation

The researchers validated EAQCO using a physical mesh network of 7 BeagleBone Black nodes. Connectivity was simulated via Ethernet using a Python-based Sockets API with UDP. Nodes generated 2.4 KB messages (simulating 100 16-byte floating-point readings) and were tested under varying packet rates to observe the impact of congestion on offloading decisions.

Results

  • Latency Reduction: In high-traffic scenarios, EAQCO significantly outperformed static routing, reducing total message processing time from 0.25 seconds to approximately 0.15 seconds.
  • Energy Conservation: The energy-aware version of the protocol consumed roughly 40% less energy than the version without energy awareness. This is attributed to nodes preferring local computation over energy-intensive multi-hop transmissions as their batteries drained.
  • Learning Dynamics: The study found that a learning rate of $\eta = 0.5$ helped the system recover faster from local computation queue spikes, while $\eta = 0.1$ was better for stable forwarding.

Conclusion and Future Work

EAQCO demonstrates that integrating computational offloading into routing protocols can significantly enhance the efficiency and longevity of WSNs. However, the protocol is currently limited to "severable" data that can be processed in independent chunks. Future work will explore more complex Quality of Service (QoS) metrics and evaluate the protocol under more realistic wireless conditions, including signal fading and variable transmission power.

Summary updated

Abstract

Wireless sensor networks (WSN) are characterized by a network of small, battery powered devices, operating remotely with no pre-existing infrastructure. The unique structure of WSN allow for novel approaches to data reduction and energy preservation. This paper presents a modification to the existing Q-routing protocol by providing an alternate action of performing sensor data reduction in place thereby reducing energy consumption, bandwidth usage, and message transmission time. The algorithm is further modified to include an energy factor which increases the cost of forwarding as energy reserves deplete. This encourages the network to conserve energy in favor of network preservation when energy reserves are low. Our experimental results show that this approach can, in periods of high network traffic, simultaneously reduce bandwidth, conserve energy, and maintain low message transition times.
BibTeX
@misc{barker2020energy,
  author = {Adam Barker and Martin Swany},
  title = {{Energy Aware Routing with Computational Offloading for Wireless Sensor Networks}},
  year = {2020},
  eprint = {2011.14795},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2011.14795},
}
Figure