Long Short Term Memory Network-based Interference Recognition for Industrial Internet of Things

Abstract

Nowadays, the rapid growth of wireless Internet of things (IoT) devices is one of the significant factors leading smart systems in various sectors, such as healthcare, education, and agriculture. This is, of course, not limited to the industrial sector, where the IoT concept is applied for real time monitoring and control of devices instead of human beings. Co-channel interferences occurs when two or more devices are using the same channel. It causes unnecessary contention as the devices will be forced to defer transmissions until the medium is clear causing a loss of throughput. Adjacent channel interference is even more serious and occurs when the devices are on overlapping channels causing corrupted data, which makes indispensable retransmissions. The more devices are added to an environment, the higher the likelihood of interference problem is. Due to a huge number of IoT devices, the interference issue becomes very serious. In this paper, a long short-term memory network-based interference recognition (LSTM-IR) is proposed. This method is integrated into the industrial IoT (IIoT) network in factory environments to mitigate the effect of interferences. The comparative results are done among three interference suppression techniques (IST) including the traditional minimum mean square error (MMSE) approach, the multi-layer perceptron (MLP), and the proposed LSTM-IR. Since the type of transmitting and receiving data is usually a sequencing data type. Therefore, the proposed method with the input data from a fast Fourier transform (FFT) algorithm provides better performances because it is based on an LSTM which is suitable for the sequences of data. The number of the devices in the factory is obviously the key factor because the smaller number of active devices causes less interferences.