LSTM based Predictive Congestion Aware Routing for Energy Efficient Wireless Sensor Networks

Abstract

Wireless Sensor Networks (WSNs) suffer from congestion due to dynamic traffic patterns, limited buffer capacity, and energy constraints, which significantly degrade packet delivery performance and network lifetime. To address these challenges, this paper proposes an LSTM-based Predictive Congestion Aware Routing framework that proactively forecasts congestion states and dynamically adapts routing decisions to enhance energy efficiency and Quality of Service (QoS). The proposed model utilizes time-series network parameters, including queue length, packet arrival rate, residual energy, and channel utilization, to train a Long Short-Term Memory (LSTM) network capable of predicting congestion levels before buffer overflow occurs. Simulation results demonstrate that the proposed approach achieves a congestion prediction accuracy of 95.8%, with a precision of 94.6% and recall of 96.2%, significantly outperforming conventional routing protocols. The high precision ensures minimal false congestion alarms, while the superior recall effectively detects critical congestion scenarios, reducing packet loss and retransmission overhead. Consequently, the model improves packet delivery ratio by 12–15%, reduces end-to-end delay by approximately 18%, and enhances overall network lifetime by nearly 20% compared to traditional reactive routing schemes. By integrating predictive intelligence with energy-aware routing metrics, the proposed framework provides a scalable and adaptive solution for next-generation WSN and IoT applications requiring reliable, low-latency, and energy-efficient communication.