DP-DTL Based Resource Allotment for Expansive and Dynamic M-WiFi/LiFi Networks

A mixed wireless fidelity (Wi-Fi)/light fidelity (Li-Fi) communication paradigm has appeared as an innovative solution to meet the requirements of contemporary and forthcoming generations of mobile systems, promising enhanced data rates and uninterrupted connectivity. Efficient provisioning of resources in these networks poses a non-convexity challenge, which has been effectively tackled through algorithms based on Deep Q-Network (DQN) learning. Nonetheless, DQN learning encounters limitations in expansive and dynamic mixed Wi-Fi/Li-Fi configurations. This study introduces an inventive dual-phase deep transfer learning (DP-DTL) approach for optimizing resource allocation in a large and dynamic mixed Wi-Fi/Li-Fi communication environment. Using transfer learning enables swift acquisition of information about newly in-troduced mobile users, significantly reducing the required iterations and enhancing the overall system throughput. The simulation results showcase the superiority of the proposed mechanism, surpassing the existing DQN learning-based approach by up to 80% in terms of the iteration count required to attain the maximum sum-rate.