RESOURCE ALLOCATION OPTIMIZATION IN IEEE 802.11AX FOR ENHANCED INDUSTRIAL INTERNET OF THINGS APPLICATIONS LEVERAGING DEEP REINFORCEMENT LEARNING

Abstract

The Internet Things (IoT) is set to multiply the number of connected gadgets tremendously and the current Wireless Fidelity (Wi-Fi) technologies face huge challenge of handling the immense flood of data. In contrast to earlier Wi-Fi standards, the IEEE 802.11ax protocol achieves higher transmission rates, better scalability, and improved coexistence, and has the added advantage of being lower cost than 5G, which makes it even more relevant in Industrial IoT (IIoT) application where multitudes of devices need to coexist in dynamic and challenging environments. Such environments are distinguished by physical barriers, heavy saturation of devices, electromagnetic static, and continual alterations in equipment as well as layout. The conventional collision avoidance mechanisms deployed by the IEEE 802.11ax networks encounter scalability challenges that affect the throughput and high delays as the network adds more devices. The optimization of the resource allocation of the Medium Access Control (MAC) layer parameters must address these challenges to support the feature-rich IIoT environments and support the realization of the requirements of heterogeneous industrial traffic and harsh factory environment. The present paper provides an optimization approach to IEEE 802.11ax systems in order to improve their functionality in IIoT. As a strategy, it involved deep reinforcement learning in order to identify the best parameter settings across different network parameters. The network was trained with parameters, which simulated an IIoT environment. The efficiency of the offered mechanism was tested with the help of MATLAB mass simulations that simulated and interpreted the behaviour of the 802.11ax network in industrial environments. The findings were that throughput on optimized network was significantly higher than unoptimised network, packet retransmission rate was lower in optimised network, and latency been lower on optimised network. Peak throughput went up by 53.07 percent to 2150 Mbps, whereas average throughput went up 39 percent. The greatest packet loss ratio reduced by 29.2% or 32.5 to 23, and the overall average packet loss was reduced by 19.7 percent. Latency also improved significantly with 5 nodes experiencing latency beyond 0.1 second narrowing to just 1 node. The results indicated that the suggested optimization approach to 802.11ax systems could improve the performance by a fair margin in IIoT setups and make them robust enough to support industrial operations.