Machine Learning Approach for Optimization of Quality of Service in Self Organizing Heterogeneous Wireless Network

Abstract

Autonomous wireless networks (AWNs) face significant challenges due to interference arising from their diverse nature and high density. Employing machine learning as an effective data-driven approach presents a promising avenue for automatic power configuration and related settings. This study outlines an innovative technique for modeling a more densely populated network by incorporating femto or pico cells, simultaneously addressing power optimization challenges through a novel reward function in a distributed network. The methodology integrates particle swarm optimization to ensure optimal parameter selection within the reward function, thereby guaranteeing the fundamental Quality of Service (QoS) for microcell users with minimal power requirements. The proposed distributed power allocation method, rooted in Q-learning, undergoes evaluation against Markov decision states. Results demonstrate that the PSO-optimized solutions outperform the greedy algorithm in achieving superior outcomes.