Cell Outage Compensation Using Q-learning for Self-Organizing Networks

Abstract

In this paper, we introduce a Q-learning-based algorithm for Cell Outage Compensation (COC) in Self Organizing Networks (SONs). The algorithm compensates the coverage in the outage area by modifying the power and antenna tilt angle parameters of the neighboring cells. The proposed Q-learning algorithm adapts the reward via learning the consequences of the taken actions to compensate the coverage gap, which guarantees a fully autonomous and accurate COC as we do not assume the knowledge of the propagation model or other models of the environment. This contrasts with existing COC approaches which are inaccurate as they assume the knowledge of the mathematical models of the system and solve the COC problem given such mathematical models. Simulation results show a 92% accessibility of the proposed Q-learning algorithm compared to 81% accessibility of existing approaches that are based on modelling the environment.