Resource Management in Next Generation Wireless Networks: Optimization and Games.

摘要

To attain the targeted data rates for next generation (5G) cellular networks, one effective approach is dense deployment of small cells, in addition to macro cells which provide wide coverage. Today's systems allocate fixed resources to a cell regardless of traffic conditions in other cells. In the inter-cell interference coordination (ICIC) scheme of current 4G networks, different resource blocks are assigned to cell-edge user equipments (UEs) from neighboring cells to mitigate inter-cell interference. However, access points cannot effectively use their neighbors' resources when there are no cell-edge UEs in the neighboring cells. Dynamic radio resource management (RRM) is crucial to the success of the heterogeneous networks due to much more pronounced traffic and interference variations in small cells.;Exploiting new licensed and unlicensed frequency bands is another promising 5G technique. Facing the challenge of providing sufficient network capacity for wireless data, the industry is currently debating how to take advantage of the hundreds of megahertz of unlicensed spectrum. As 4G wireless technology, Long Term Evolution (LTE) has benefits of high-efficiency and robust mobility in comparison with WiFi. One specific proposal being considered by the 3rd Generation Partnership Project (3GPP) is to retool and deploy LTE technologies in unlicensed bands below 6 GHz. Along with benefits, challenging issues, such as coordination among operators and coexistence among radio access technologies, need to be carefully addressed.;The first part of this thesis addresses resource management in dense small cells. In Chapter 2, we propose a framework for RRM organized according to two timescales, which includes 1) centralized RRM on a relatively slow timescale (seconds to minutes) corresponding to typical durations of user sessions, and 2) distributed RRM on a very fast timescale (milliseconds) corresponding to typical latency requirements. We treat UEs near each other with similar quality of service (QoS) requirements as a UE group and model their aggregate traffic using a single queue on a slow timescale. In reality, the service rate for one UE group depends on spectrum activities of other UE groups. The network becomes a system of interactive queues. The analysis of interactive queues is an open problem. To make progress, we propose an effective approximation of the average packet delay. The joint design of dual timescale allocation and user association is studied. An optimization problem to minimize average packet delay is formulated with consideration of (distributed) opportunistic scheduling on the faster timescale. Iterative algorithms are developed to solve the optimization problems efficiently for a small cluster of cells. Numerical results demonstrate advantages of the dual-timescale RRM framework.;Chapters 3 and 4 study the fundamental questions of whether and how the unlicensed spectrum can be shared by intrinsically strategic operators with dynamic traffic. First, we propose a static sharing scheme, which allows the operators to share the spectrum over a slow timescale and has a subgame-perfect Nash equilibrium. The question of how many operators will choose to enter the market is also addressed by considering an entry game. Several dynamic sharing schemes are then proposed, where borrowing/lending is allowed among operators. The efficiency of these dynamic schemes significantly outperforms that of static sharing, and the strategy profiles in these sharing schemes are also shown to be subgame-perfect Nash equilibria. Chapter 3 shows the existence of the subgame-perfect Nash equilibria when the future discount is sufficiently close to 1. Chapter 4 studies the efficient subgame-perfect Nash equilibria to maximize social welfare for general cases. Practical implication of the results to the deployment of LTE in unlicensed bands are discussed.;In Chapter 5, the coexistence issue between LTE in unlicensed spectrum (LTE-U) and other technologies via channel selection is discussed. A simple case with multiple LTE-U operators and one foreigner using a different technology is discussed here. In practice, the foreigner may be either strategic or nonstrategic. For different types of foreigners, the optimal strategy of LTE-U operators is discussed using game theoretic tools.;The study in Chapters 3, 4 and 5 is to provide a basis for reasoning of the deployment of LTE in unlicensed bands. The choice of the actual equilibrium allocation is likely to be determined by standard bodies (including the 3GPP and WiFi Alliance) as well as other major parties holding a stake in the outcome.