Self-optimized heterogeneous networks for energy efficiency

Abstract

Abstract Explosive increase in mobile data traffic driven by the demand for higher data rates and ever-increasing number of wireless users results in a significant increase in power consumption and operating cost of communication networks. Heterogeneous networks (HetNets) provide a variety of coverage and capacity options through the use of cells of different sizes. In these networks, an active/sleep scheduling strategy for base stations (BSs) becomes an effective way to match capacity to demand and also improve energy efficiency. At the same time, environmental awareness and self-organizing features are expected to play important roles in improving the network performance. In this paper, we propose a new active/sleep scheduling scheme based on the user activity sensing of small cell BSs. To this end, coverage probability, network capacity, and energy consumption of the proposed scheme in K-tier heterogeneous networks are analyzed using stochastic geometry, accounting for cell association uncertainties due to random positioning of users and BSs, channel conditions, and interference. Based on the analysis, we propose a sensing probability optimization (SPO) approach based on reinforcement learning to acquire the experience of optimizing the user activity sensing probability of each small cell tier. Simulation results show that SPO adapts well to user activity fluctuations and improves energy efficiency while maintaining network capacity and coverage probability guarantees.