⚡ electrical engineering

An Energy Harvesting and Prediction-Driven Routing Protocol for Long-Range Linear Wireless Sensor Networks

This paper proposes an Energy-Time-Aware Routing Strategy (ETARS) guided by a novel Enhanced Weather Conditional Moving Average (E-WCMA) predictor to optimize relay selection and mitigate energy shortages, queue congestion, and information staleness in long-range linear energy-harvesting wireless sensor networks.

Original authors: Haibo Yang, Hongguang Xiao, Weimin Lei, Junying Jia

Published 2026-09-01

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Haibo Yang, Hongguang Xiao, Weimin Lei, Junying Jia

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

In the vast, open corridors of the world, from the steel ribs of a railway line to the winding paths of a pipeline, a quiet revolution is taking place in how we watch over our infrastructure. For decades, engineers have relied on wireless sensor networks to monitor these long, linear stretches of land, placing hundreds of small, battery-powered devices along the route to gather data on temperature, vibration, and structural health. However, these devices face a stubborn, physical reality: they run on finite batteries. In remote locations where maintenance crews cannot easily reach to swap out power cells, a single dead battery can create a blind spot, severing the chain of information and leaving critical infrastructure unmonitored. To solve this, scientists have turned to energy harvesting, a method where sensors scavenge power from their surroundings, such as sunlight or wind, to recharge their batteries indefinitely. Yet, nature is unpredictable. The sun hides behind clouds, and the wind dies down, creating a chaotic rhythm of energy availability that makes it difficult for a sensor to know if it has enough power to send its next message. If a sensor guesses wrong, it might try to transmit when its battery is empty, or worse, it might sleep when it should be awake, causing the information it holds to grow stale and useless by the time it finally reaches the central computer.

A team of researchers from Shenyang University of Technology and Northeastern University has developed a new approach to keep these long-range sensor networks alive and their information fresh. They recognized that simply knowing how much energy a sensor has right now is not enough; the network needs to know what the weather will bring in the next few hours. To achieve this, they created a smarter way to forecast energy availability, one that learns from past mistakes and adapts to sudden changes in the environment. They paired this forecasting tool with a new set of rules for how the sensors talk to each other. Instead of just passing data to the nearest neighbor, the network now makes decisions based on a combination of three factors: how much energy a neighbor is predicted to have, how many messages are waiting in that neighbor's queue, and how far the message needs to travel. This strategy ensures that data flows through the healthiest, least congested paths, keeping the network connected even when individual sensors run low on power.

The researchers tested their ideas using real-world data collected from a solar-powered testbed and simulated the behavior of a fifty-node network along a railway corridor. Their forecasting method, which they call an enhanced weather conditional moving average, proved to be more accurate than existing techniques. By adjusting its internal settings based on how wrong it was in the previous moments, the system learned to weigh recent weather patterns more heavily than older ones, allowing it to react quickly to sudden cloud cover or clearing skies. In their simulations, this improved prediction allowed the network to make better choices about which sensor should forward a message. When the researchers compared their new routing strategy against older methods, they found that their approach preserved significantly more energy in the network's batteries over time. More importantly, it prevented the buildup of data backlogs. In the older systems, messages would pile up at certain nodes, waiting for a chance to be sent, which caused the information to age rapidly. The new system, by prioritizing nodes with fewer waiting messages, kept the flow of information steady and reduced the time it took for fresh data to reach the destination.

A critical part of their design addresses the problem of "routing holes," which occur when a sensor in the middle of the line runs out of energy and cannot pass messages to the next layer. In many traditional networks, this single failure would break the entire chain, stopping all communication from the sensors behind it. The new strategy includes a safety mechanism that allows a sensor to send a message sideways to a neighbor on the same level if the path forward is blocked. This lateral hop is a temporary detour, designed only to find a new starting point to jump back down the chain toward the central receiver. The simulations showed that this recovery step was essential; without it, the network would disconnect much earlier in its operational life. By combining accurate energy prediction with a flexible routing plan that considers both battery levels and data queues, the researchers demonstrated that it is possible to maintain a reliable, long-term monitoring system in environments where power is scarce and unpredictable. Their work suggests that with the right algorithms, these sensor networks can operate for years without human intervention, providing a continuous, real-time view of the infrastructure that keeps our world moving.

Technical Summary: An Energy Harvesting and Prediction-Driven Routing Protocol for Long-Range Linear Wireless Sensor Networks

Problem Statement
Wireless Sensor Networks (WSNs) deployed in long-range linear scenarios, such as railway monitoring, face a critical bottleneck: limited battery power combined with restricted maintenance access. While Energy Harvesting (EH) technology offers a solution by replenishing energy from environmental sources (e.g., solar), the stochastic and time-varying nature of energy harvesting introduces significant uncertainty. This uncertainty leads to energy shortages, queue congestion, and increased Age of Information (AoI), where data becomes stale before reaching the sink. Existing routing protocols often rely on instantaneous residual energy or fixed prediction models, failing to adequately adapt to rapid weather changes or the dual energy costs of data collection and forwarding in linear topologies. Consequently, these methods struggle to maintain route continuity and information freshness under late-stage energy scarcity.

Methodology
The paper proposes a comprehensive framework consisting of two main components: an enhanced energy prediction algorithm and a prediction-driven routing strategy.

Enhanced Weather Conditional Moving Average (E-WCMA):
To address the limitations of fixed-parameter statistical predictors, the authors designed E-WCMA. This algorithm retains the lightweight structure of the traditional Weather Conditional Moving Average (WCMA) but introduces three specific improvements:
- Adaptive Parameter Adjustment: Parameters ( $\alpha$ , $D$ , and $K$ ) are dynamically updated based on the deviation between instantaneous prediction errors and a sliding window of historical errors. This allows the model to increase historical weighting during abrupt fluctuations and prioritize recent data during stable periods.
- Exponentially Weighted Historical Mean: Instead of equal weighting for historical samples, the algorithm applies exponential decay, giving more influence to recent days to better capture local seasonal or shading changes.
- Enhanced GAP Factor: A weighted ratio is computed between current observations and historical trends, allowing the correction factor to be dimensionless and more responsive to recent deviations.
Energy-Time-Aware Routing Strategy (ETARS):
ETARS is a multi-objective routing protocol designed for layered linear topologies. It selects relays based on a composite score derived from three normalized metrics:
- Available Energy: A combination of current residual energy and predicted harvestable energy (derived from E-WCMA).
- Queue Backlog: The number of packets waiting in the node's buffer, serving as a proxy for information freshness pressure.
- Forwarding Distance: The Euclidean distance to the next hop.
The weights for these metrics are determined via pairwise comparison, prioritizing queue length (0.50) to prevent data stagnation, followed by available energy (0.33), and finally distance (0.17).
- Intra-Layer Forwarding: To handle "routing holes" where no feasible lower-layer relay exists (due to energy depletion), the protocol includes a recovery mechanism. If downward forwarding fails, the node attempts to forward the packet laterally to an unvisited same-layer neighbor that meets a residual energy threshold. This is a conservative recovery step used only when normal downward paths are unavailable.

Key Contributions

E-WCMA Algorithm: A lightweight, adaptive prediction algorithm that improves accuracy and stability under complex weather conditions without requiring additional hardware or heavy computational resources.
ETARS Protocol: A routing strategy that jointly optimizes energy sustainability, traffic load (via queue awareness), and forwarding efficiency. It explicitly integrates predicted energy into the relay selection metric.
Local Recovery Mechanism: The introduction of an intra-layer forwarding step to maintain route continuity in linear deployments when lower-layer relays become unavailable, preventing network partitioning.

Experimental Results
The study utilized a public solar energy harvesting dataset for prediction validation and OMNeT++ simulations for routing evaluation.

Prediction Performance: On the testbed dataset, E-WCMA achieved the lowest error rates across all metrics (MAPE, WMAPE, MAE, RMSE) compared to baseline algorithms (Pro-Energy, WCMA, EWMA). Specifically, it reduced WMAPE by 9.92% and RMSE by 11.27% compared to the best-performing baseline.
Routing Performance: In simulations, ETARS demonstrated superior performance in preserving residual energy and maintaining lower AoI compared to ERMRM, MAX-SNR, and ORDPD.
- Energy & Queue: ETARS maintained lower queue accumulation and higher residual energy in the late stages of network operation (beyond round 25), whereas baselines suffered from rapid energy depletion and queue growth.
- AoI: ETARS achieved the lowest average peak AoI (approximately 33% reduction compared to ERMRM) by prioritizing nodes with high backlog pressure.
- Robustness: Under increasing packet loads, ETARS maintained route continuity while baseline schemes experienced routing disruptions and flat AoI traces (indicating a cessation of fresh data delivery).
Ablation Studies: Removing the prediction component or the queue-priority weighting resulted in earlier network disruption and higher energy consumption in the late stages, confirming the necessity of both accurate prediction and queue-aware routing.

Significance and Claims
The paper claims that the proposed framework effectively addresses the trade-off between energy sustainability and information timeliness in linear EH-WSNs. By integrating adaptive energy prediction with a multi-objective routing strategy, the system can better anticipate energy availability and mitigate queue-induced delays. The authors state that this approach has the potential to reduce the maintenance burden of long-range sensing infrastructure and provides a reference for green monitoring deployments in extended corridor scenarios. The work is presented as a modest improvement over existing static or instantaneous-state-based methods, specifically tailored to the constraints of linear topologies and the limitations of low-power sensor nodes.

Technical Summary: An Energy Harvesting and Prediction-Driven Routing Protocol for Long-Range Linear Wireless Sensor Networks

More like this