i5GAccess: Nash Q-learning Based Multi-Service Edge Users Access in 5G Heterogeneous Networks

In the heterogeneous wireless networks, it remains a significant challenge to achieve an efficient network selection strategy to satisfy the demands of a massive number of edge users and novel 5G services. In this paper, we formulate the network selection problem for edge users as a discrete-time Markov model, and propose a Nash Q-learning based intelligent network access algorithm for multi-agent system, named MAQNS. We consider the joint optimization of network selection strategies among different types of networks, aiming at maximizing the long-term performance of multi-agent system. Meanwhile, we use Analytic Hierarchy Process (AHP) and Grey Relation Analysis (GRA) to characterize the user preferences for networks. Experimental results show that comparing to the existing network selection algorithms, the proposed MAQNS has better performance in terms of system throughput, user blocking probability, average energy efficiency and average delay.