Deep Learning-based LOS/NLOS Classification for Reliable Communication in Industrial IoT

The optimal channel selection for data transmission is essential for reliable communication in an indoor environment like industrial IoT (IIoT). Due to the presence of complex objects in the indoor factory environment, signals might get reflected. This leads to reliability loss and degradation of transmitted signal quality by increasing signal outage. Again, the optimal channel selection for efficient scheduling of the heterogeneous data packets generated by delay-sensitive ultra-reliable low latency (URLLC) service and delay-tolerant broadband service in IIoT demand for accurate identification of wireless link status. Therefore, the identification of wireless channel status like Line-of-Sight (LOS), None-Line-of-Sight (NLOS), and Multi-Path (MP) between the transmitter and the receiver becomes essential to reduce packet error probability in IIoT. In this regard, we propose a deep learning-based classifier model using Convolutional Neural Network (CNN) and Long Short Term Memory (LSTM) to identify the LOS/NLOS or MP signals and enhance the reliability of the signal by improving system throughput and accurate positioning. We compare the performance of the proposed model with various machine learning classifier models to evaluate the performance of the proposed CNN-LSTM model. We have used an open-source dataset collected from two different indoor industrial sites to be used for the training and testing of the classifier models. We have evaluated the performance based on the accuracy and time complexity of the proposed classifier model, which shows superiority in comparison to baseline machine learning models. Additionally, the results show that the system Bit Error Rate (BER) improved significantly with the optimal channel selection during scheduling of heterogenous data type using the proposed CNN-LSTM model.