Efficient User Association using Multi-Objective Q-Learning Approach

User association is a critical issue for the Heterogeneous Networks(HetNet) due to a number of criteria, such as resource allocation, mitigation of interferences, and increased data rate. Various works have been proposed for the efficient user joint as the KNN model, where the users are associated with the nearest BS. However, the majority of the previous works do not study the big-data network scenarios. In this work, we propose a multi-objective user association using Q-Learning algorithm by considering Data Rate, Energy Efficiency (EE), and load balancing simultaneously. The new method learns optimal associations based on the reinforcement learning tool, avoiding BS overload while enhancing energy efficiency and maintaining stable SINR. simulations results prove that the Q-Learning-based association outperforms KNN in terms of average Data Rate and EE.