Quality-Aware Data Aggregation in Environmental WSNs Using Autoencoder Filtering

Abstract

Environmental monitoring using Wireless Sensor Networks (WSNs) requires accurate and reliable data to support real-time decision-making. However, traditional data aggregation techniques often suffer from issues such as redundancy, noise, outliers, and unreliable readings, which compromise data integrity. This paper proposes a hybrid quality-aware data aggregation method that integrates autoencoder-based anomaly detection with rule-based filtering to enhance data fidelity. The autoencoder identifies anomalies through reconstruction error analysis, while the rule-based system applies sensor-specific thresholds to eliminate invalid values. The approach was implemented and simulated in MATLAB using real-world sensor data patterns. Performance was evaluated against conventional LEACH, PEGASIS, K-LEACH, and MABRL protocols based on metrics such as aggregation accuracy, packet delivery ratio, data loss rate, routing overhead, and end-to-end delay. Results show that the proposed method consistently achieves higher accuracy across various sensor types and significantly improves network reliability and communication efficiency. It also reduces unnecessary transmissions, leading to lower energy consumption and extended network lifespan. This work presents a robust solution for ensuring high-quality data aggregation in environmental WSN applications, contributing to more accurate and sustainable environmental monitoring systems.