Predictive association control in dense femtocell networks

Abstract

The deployment of femtocells with small radius can cause a rapid rise in the number of initiated handovers since users stay in the femtocells for relatively short time. And frequent handovers result in the increase of signaling overhead for handover management. As a result, quality of services decreases due to handover interruptions. For these reasons, the elimination of unnecessary handovers or the minimization of frequent handovers is a critical problem in the femtocell networks. However, conventional approaches fail to appreciate how frequent handovers affect the throughput performance of a moving mobile station in femtocell networks. In addition, problems relating to creating and managing the neighbor cell lists have a considerable effect on handover performance. In this thesis, we propose two predictive base station (BS) association control schemes for femtocell networks, based on uplink and downlink models, respectively. In the uplink based association control scheme, we suggest a predictive scheme to generate reduced neighbor cell lists. In addition, a predictive association game is formulated. The game is designed to determine an optimal base station for maximum throughput and minimized unnecessary handovers. In this non-cooperative game, each mobile station independently chooses its transmission power and base station association through a composite cost function in order to maximize throughput and to reduce unnecessary handover. In the downlink based association control scheme, we consider the association control problem of the mobile station as an optimal sequential decision problem and obtain a rolling-out suboptimal control policy which is based on the predictive mobility of the mobile station. In addition, we derive a simple and lightweight lookahead association control policy for each mobile station. Several simulation scenarios are set forth herein for analysis to show how the proposed schemes improve performance in terms of throughput and han...