Utility-based cross-layer radio resource management in wireless networks

Abstract

In this thesis, we consider various types of cross-layer resource allocation problems in the context of long-term network utility maximization. This utility-based cross-layer radio resource management we are touching on covers from single-hop single-carrier systems up to multi-hop multi-carrier systems. In Chapter 2, we consider single-hop single-carrier systems and propose a combined scheduling and call admission control algorithm that exactly guarantees the minimum throughput requirements of CBR (constant bit rate) and EMG (elastic with minimum guarantee) users, and then allocates the leftover capacity to EMG users in proportional fair manner. We also show that by establishing priority relationship in the scheduling, the proposed scheduling algorithm can be naturally used for measurement-based call admission control, which does not adversely affect ongoing calls. In Chapter 3, we propose a joint subcarrier and power allocation achieving asymptotic network utility maximization (NUM). Using the proposed algorithm, we revisit equal power allocation and show that the equal power allocation, which is known to be near optimal to sum capacity maximization problem in high SNR regime, can produce significantly wrong answer to general NUM problem with various SNR distributions. In particular, equal power allocation shows poor performance under the resource sharing policy which strongly emphasizes fairness like max-min. Our focus moves onto the cross-layer design in multi-hop wireless networks. In Chapter 4, we consider achieving long-term NUM in multi-hop wireless networks. To this end, we develop a joint flow control and resource allocation algorithm for a priori fixed routing path. This does not impose any specific physical layer model so that it is general enough to be applied to any model. Further, we propose an iterative method that computes scheduling decision in multi-hop single-carrier networks. We show that the proposed method attains an optimal solution...