RL-MAB-Based Resource Allocation for Efficient Bandwidth Utilization in Industrial IoT Networks

Abstract

Industrial IoT (IIoT) is key to achieving automation in manufacturing and logistics industries. Recently, there have been ongoing efforts to realize automation by deploying mobile devices such as Automated Guided Vehicles (AGVs) in industrial processes, along with sensors and actuators. Industrial Wireless Sensor Networks (IWSN) are essential foundational technologies for implementing IIoT, providing communication methods based on redundancy policies, such as scheduling with pre-designed multiple routing paths and fixed redundant resource allocation, to meet strict industrial requirements like real-time, reliability, and energy efficiency. However, existing IWSNs are vulnerable to frequent location changes of equipment such as AGVs and variations in network environments. Especially, the waste of redundant resources that are not utilized for communication leads to issues of resource efficiency and scalability limitations due to limited wireless resources. To address these challenges, this paper proposes a resource reallocation and scheduling update scheme that identifies and reallocates unutilized redundant resources by employing the Multi-Armed Bandit (MAB) algorithm-one of the representative reinforcement learning techniques focused on reward optimization-in IWSN environment where mobile and fixed devices are deployed together. This approach improves the performance of industrial networks while maintaining compatibility with existing industrial network systems, supporting reliability and real-time processing. The experimental performance evaluation shows that the proposed scheme improves network performance in terms of resource efficiency, scalability, energy efficiency, and mobility support compared to the typical resource allocation and scheduling schemes.