Deep learning based optimal restricted access window mechanism for performance enhancement of IEEE 802.11ah dense IoT networks

IEEE 802.11ah, primarily developed for Internet of things (IoT) applications. A restricted access window (RAW) mechanism is employed for channel access to minimize contention among the nodes in dense IoT networks. Here, we present an analytical model for IEEE 802.11ah RAW mechanism with unsaturated conditions. In IEEE 802.11ah, determination of optimal number of RAW slots/groups affects the network performance. Thus, choosing the optimal number of RAW slots according to network size, RAW duration and modulation coding scheme (MCS) can improve the performance of IEEE 802.11ah network. Recently, deep learning (DL) becomes a thriving field for communication and networking applications. In this research work, we propose a long short-term memory (LSTM) based DL-recurrent neural network (DL-RNN) to predict the optimal RAW slots. Using genetic algorithm, we obtain optimal RAW slots from the presented analytical model with unsaturated conditions for the IEEE 802.11ah IoT network. Further, the dataset with optimal number of RAW slots for each of the combinations (network sizes, RAW durations and MCSs) is used for training the LSTM based DL-RNN. Based on the optimal RAW slots, throughput and energy efficiency (EE) are computed for IEEE 802.11ah network. From results, we notice that optimal RAW slots generated by LSTM based DL-RNN model significantly improve throughput and EE of IEEE 802.11ah network comparison to existed machine learning and Fuzzy optimization techniques.