Hybrid deep learning framework for RSSI and SNR classification in industrial LoRaWAN networks

Abstract Reliable wireless communication is critical for Industry 4.0, but LoRaWAN performance degrades in harsh industrial environments due to electromagnetic interference and signal fluctuations. Accurately classifying key indicators like Received Signal Strength Indicator (RSSI) and Signal-to-Noise Ratio (SNR) is a major challenge. Existing methods, including standalone deep learning models, often fail to capture both the complex feature interactions and the crucial temporal dependencies of these signals. To overcome these limitations, this study proposes TabLSTMNet, a novel hybrid deep learning model. TabLSTMNet uniquely combines TabNet’s adaptive feature selection with LSTM’s robust temporal modeling, allowing it to efficiently analyze and classify dynamic RSSI and SNR data. Trained on a large-scale industrial LoRaWAN dataset, the model demonstrates superior accuracy and robustness compared to conventional and individual deep learning approaches. This work contributes to the optimization of industrial wireless networks by enabling more reliable IoT communication, proactive network management, and enhanced resilience for smart industrial systems.