AI-Powered Dynamic Resource Allocation Approach for Improving Utilization in High-Speed 5G Wireless Communication Networks
TL;DR
A novel hybrid RA scheme was designed by combining the strengths of a recurrent neural network (RNN) and a deep Q-network (DQN) to allocate resources optimally to users while meeting various QoS requirements.
Abstract
The introduction of the 5G wireless communication network has proclaimed a new era of internet connectivity, allowing greater data speed, lower latency, and high capacity. The efficiency of the 5G network relies on optimal resource allocation (RA), which determines how the available computational resources are prudently allocated to user equipment and devices. Although various RA schemes have been developed, they rely on predefined protocols and cannot be optimal at all times. In this research, a novel hybrid RA scheme was designed by combining the strengths of a recurrent neural network (RNN) and a deep Q-network (DQN). The proposed model aims to allocate resources optimally to users while meeting various QoS requirements. In the proposed scheme, initially, data from the multi-channel wireless system are accumulated and pre-processed to enhance the training process. The integrated RNN function learns the temporal interconnections between the input parameters and generates a resource allocation plan. DQN provides the Q-value and assists the RNN in making optimal RA decisions. In addition, the developed model utilizes the Krill Herd Optimization (KHO) for fine-tuning the hyperparameters of RNN-DQN, thereby enhancing the convergence speed and improving system performance. Finally, an intensive simulation analysis was performed and the efficiency of the designed model was evaluated with a comparative study.