Dynamic Adaptive Reinforcement Learning based Resource Allocation in Industrial Internet of Things

The resource allocation in distributed environment issue in energy efficient information to ward in resource constraint Industrial Internet of Things (IIoT). Though, because of lacking the flexibility, classical duty cycling mechanism causes addition collision of information and re-transfer particularly in challenging and dynamic environments, resulted in high energy consumption. To overcome these drawbacks, the Dynamic Adaptive Reinforcement Learning (DARL) algorithm for allocating the resources in IIoT with less energy consumption. The dynamic adaptive process is included in the process of RL algorithm, which enhances the performance of RL for resource allocation. The RL method is the dynamic program that addressed the issue through formulating the optimization issue of Multiple Agent RL (MARL), where every Base Station (BS) considered as the learning agent. The developed DARL algorithm evaluates the First Node Death Time (FNDT) in terms of number of nodes. The developed DARL reached 12s on 20 nodes, 14s on 40 nodes, 19s on 60 nodes, 20s on 80 nodes and 23s on 100 nodes while comparing with Adaptive RL (ARL) algorithm.