Title Page
Abstract
Contents
List of Acronyms 19
Chapter 1. Introduction 21
1.1. Background 21
1.2. Motivations 24
1.3. Dissertation organization 26
Chapter 2. State of the Art 28
2.1. Overview 28
2.1.1. Bit loading 30
2.1.2. Wireless schedulers 32
2.1.3. Throughput analysis of schedulers 34
2.2. Limitations 35
2.2.1. Limitations in ABL 35
2.2.2. Limitations of existing wireless schedulers 37
2.3. Contributions 40
Chapter 3. Resource Allocation in OFDMA based Systems 42
3.1. General description 42
3.2. Adaptive bit loading 43
3.3. Sub-carrier scheduling and power allocation 46
Chapter 4. Adaptive Grouping and Weighted Loading for MIMO OFDM Systems 54
4.1. System model 55
4.2. Weighted loading 58
4.3. Adaptive grouping 61
Chapter 5. Weighted Normalized User's CSI based Scheduler 67
5.1. System model 68
5.2. Opportunistic schedulers 70
5.2.1. Scheduling algorithms 70
5.2.2. Weighted PFS: weighted normalized user's CSI based scheduler 74
5.3. Throughput analysis of weighted PFS 79
5.3.1. Single path Rayleigh fading channel 82
5.3.2. Multi-path Rayleigh fading channel 87
Chapter 6. Performance Evaluation 96
6.1. Performance evaluation for weighted loading and AG algorithm 96
6.1.1. Performance of weight factor in bit loading algorithms 98
6.1.2. Performance of adaptive grouping algorithm 100
6.2. Performance evaluation of weighted normalized user's CSI based scheduler 107
6.2.1. Complexity of the algorithms 107
6.2.2. Capacity performance of the algorithms 109
6.2.3. Proportionality maintained by the algorithms 110
6.3. Performance of weighted normalized user's CSI based scheduler with adaptive grouping 111
6.4. Performance evaluation of analytical expression for average throughput of weighted PFS 114
Chapter 7. Conclusions 122
Appendices 13
A. Proof of Lemma 2 125
B. Proof of Lemma 4 127
C. Proof of Proposition 5 128
D. Long-term average throughput of PFS 130
References 133
Table 6.1. Simulation Parameters for MIMO-OFDM system 97
Table 6.2. Simulation Parameters for OFDMA system 115
Figure 2.1. Radio resource allocation entities. 29
Figure 4.1. Block diagram for bit loaded MIMO-OFDM system. 56
Figure 4.2. Average group size and average number of groups for different limit value (δ) in GL for ITU pedestrian A channel model for the average SNR 64
Figure 4.3. Average group size and average number of groups for different limit value (δ) in GL for ITU vehicular A channel model for the average SNR 65
Figure 4.4. Example of adaptive grouping of channel gain 66
Figure 5.1. Block diagram for resource allocation in OFDMA based downlink system. 69
Figure 5.2. Demo of the proposed scheduler with K=2, N=64, and other algorithm parameters as in Algorithm 2. 78
Figure 5.3. Moving average of weights for K=4. 93
Figure 5.4. Exact vs. approximated average throughput plots. 95
Figure 6.1. Comparison of initial bit estimation. 98
Figure 6.2. Initial bit estimation of FDBA with weight factor. 99
Figure 6.3. BER curves (ITU Pedestrian A channel model). 100
Figure 6.4. BER vs. limit value(δ) (g=28.5㏈, ITU Pedestrian A channel model).(이미지참조) 102
Figure 6.5. BER curves (ITU Vehicular A channel model). 103
Figure 6.6. BER vs. limit value(δ) (g=27.6㏈, ITU Vehicular A channel model).(이미지참조) 104
Figure 6.7. Coded BER curves (ITU Pedestrian A channel model). 105
Figure 6.8. Coded BER curves (ITU Vehicular A channel model). 106
Figure 6.9. Average CPU time for sub-carrier and power allocation for various algorithms(using Matlab on Windows 7 w. Intel Core2 Duo 2.93㎓ CPU). 108
Figure 6.10. Sum capacity vs. the number of users (K). 109
Figure 6.11. Normalized capacity ratios per user, K=6 (required user rate proportion at the leftmost bar for each user). 110
Figure 6.12. Average sum capacity with full CSI, and partial CSI using (AG (δ=0.15) and UG) algorithms. 112
Figure 6.13. Normalized capacity ratios per user with full CSI, and partial CSI (AG and UG) algorithms for K=6 (required user rate proportion at the leftmost bar for each user). 113
Figure 6.14. Comparison of simulated and theoretical average throughput for weighted PFS under single path Rayleigh fading channel. 116
Figure 6.15. Comparison of simulated and theoretical average throughput for weighted PFS under multi-path Rayleigh fading channel. 117
Figure 6.16. Scheduling gain achieved by weighted PFS when compared to RRS 118
Figure 6.17. Average throughput vs. average SNR (gk=g∀k) for K=6.(이미지참조) 119
Figure 6.18. Average throughput vs. channel-taps (L) for K=6. 120