Resource Allocation in Multi-carrier Full-duplex Networks

학위논문 (박사)-- 서울대학교 대학원 : 전기·컴퓨터공학부, 2015. 2. 박세웅.Recent advances in the physical layer have demonstrated the feasibility of in-band wireless full-duplex for a node to simultaneously transmit and receive on the same frequency band. While the full-duplex operation can ideally double throughput, the network-level gain of full-duplex in large-scale networks remains unclear due to the complicated resource allocation in multi-carrier environments. In this dissertation, we tackle three different resource allocation problems in multi-carrier full-duplex networks. Firstly, we investigate the power allocation problem in three-node full-duplex OFDM networks where one full-duplex node transmits to a half-duplex node while receiving from another half-duplex node at the same time. We formulate the sum-rate maximization problem with and without joint decoding, and develop a low-complexity solution for each case. Through simulations, we evaluate our proposed solutions and demonstrate the full-duplex gain in various scenarios. Secondly, we consider the resource allocation problem in full-duplex OFDMA networks where both the base station and mobile nodes are full-duplex capable. We propose a joint solution to the subcarrier assignment and power allocation problem by establishing a necessary condition for the sum-rate optimality. We show that our algorithm is provably efficient in achieving \emph{local Pareto optimality} under certain conditions that are frequently met in practice. Through extensive simulations, we show that our algorithm empirically achieves near-optimal performance. Lastly, we investigate the resource allocation problem in full-duplex OFDMA networks where the base station is full-duplex capable while mobile nodes are conventional half-duplex nodes. Specifically, we consider two different cases where i) the BS knows all channel gains and ii) the BS obtains limited channel information through channel feedback from each node. In the former case, we design a sequential resource allocation algorithm which assigns subcarriers to uplink nodes first and downlink nodes or vice versa. In the latter case, we propose a low-overhead channel feedback protocol where downlink nodes can estimate inter-node interference by overheating feedback messages transmitted by uplink nodes. We evaluate our solutions under various scenarios through simulations.1 Introduction 1.1 Motivation 1.2 Background and Related Work 1.3 Contributions and Outline 2 Power Allocation with Inter-node Interference in Full-duplex OFDM Networks 2.1 Introduction 2.2 System Model and Problem Formulation 2.3 Power Allocation with Joint Decoding 2.3.1 Convex Problem and Dual Formulation 2.3.2 Optimal Power Allocation via Dual Optimization 2.4 Power Allocation without Joint Decoding 2.4.1 Necessary Conditions for Local Optimality 2.4.2 Optimality of FDMA Power Allocation 2.4.3 NP-hardness of Finding Optimal Power Allocation 2.4.4 Partial FDMA Power Allocation 2.5 Performance Evaluation 2.6 Summary 3 Resource Allocation in Full-duplex OFDMA Networks 3.1 Introduction 3.2 System Model 3.3 Necessary Condition for Optimality 3.4 Proposed Resource Allocation Algorithm 3.4.1 Power Allocation 3.4.2 Subcarrier Assignment 3.5 Local Pareto Optimality 3.6 Performance Bound 3.7 Performance Evaluation 3.7.1 Simulation Results 3.8 Summary 4 Resource Allocation with Inter-node Interference in Full-duplex OFDMA Networks 4.1 Introduction 4.2 System Model and Problem Formulation 4.3 Resource Allocation with Full CSI 4.3.1 Subcarrier Assignment Condition 4.3.2 Proposed Resource Allocation Algorithms 4.3.3 Asymtotic Analysis of Full-duplex Gain 4.4 Resource Allocation with Limited CSI 4.4.1 Challenge of Channel Feedback 4.4.2 Proposed Feedback Protocol 4.4.3 Calculation of Thresholds 4.4.4 Performance Analysis and Optimal Feedback Probability 4.5 Performance Evaluation 4.5.1 Simulation Setting 4.5.2 Simulation Results: Full CSI 4.5.3 Simulation Results: Limited CSI 4.6 Summary 5 Conclusion 5.1 Research Contributions 5.2 Future Research DirectionsDocto