Resource allocation and performance optimization in full-duplex MIMO/OFDMA systems

Abstract

With the development of self-interference (SI) cancellation, full-duplex (FD) radios, i.e., using the same frequency channel for transmit and receive, have recently gained significant attention owing to the potential to further improve or even double the capacity of conventional half-duplex (HD) systems. Although the gains of full-duplex systems can be easily foreseen, practical implementations of such full-duplex systems pose many challenges and a lot of technical problems still need to be solved. Moreover, many wireless systems are starting to use orthogonal frequency division multiple access (OFDMA) and multiple-input multiple-output (MIMO) as the core transmission techniques. In addition, the cooperative relaying technique is being considered to further increase the capacity of mobile cellular systems. Applying the full-duplex technology in MIMO/OFDM and/or cooperative systems will bring more degrees of freedom in system design and resource allocation, and therefore needs more insightful investigations. In this thesis, we will explore the potential of full-duplex technology at the base station (BS) in MIMO/OFDM mobile cellular systems with/without relays while the user terminals are operating in the half-duplex mode. Firstly, we consider an OFDMA multi-user cellular system with one full-duplex base station communicating with multiple half-duplex users in a bidirectional way. The uplink and downlink transmissions are coupled together due to the existence of the self-interference (SI) at the base station and the inter-user interference (IUI) from the uplink users to the downlink users. We aim to maximize the system sum-rate of uplink and downlink transmissions by optimally pairing the uplink and downlink users, and allocating the subcarriers and powers to these users. We formulate the problem as a mixed integer nonlinear programming problem. A two-layer iterative solution based on the dual method and the sequential parametric convex approximation (SPCA) method is proposed. It is referred to as the Dual-SPCA algorithm. The Dual-SPCA algorithm requires the IUI channel state information (CSI) to be available at the base station and hence a significant overhead is generated. To reduce the amount of overhead required, we assume that the IUI channel model is known at the BS and we design a location-aware resource allocation algorithm with limited CSI that maximizes the system sum-rate. Simulation results show that when SI is low, uplink and downlink user-pairing can provide significant improvement on the system sum-rate compared to the conventional unidirectional half-duplex transmission. In addition, by considering two different network deployments, i.e., urban macro cell scenario and small cell scenario, we show that the improvement of full-duplex transmission over half-duplex transmission highly depends on the channel parameters. Secondly, we jointly consider three different transmission modes in cooperative OFDMA systems, i.e., direct transmission mode, half-duplex relay cooperative transmission mode and full-duplex relay transmission mode. The joint optimization problem of transmission mode selection, subcarrier assignment, relay selection, subcarrier-pairing as well as power allocation is investigated. We transform the binary assignment problem into a maximum weighted bipartite matching problem. Based on the dual method, we solve the joint power allocation and binary assignment problem iteratively. Specifically, since the direct link is considered to be interference in the full-duplex relay transmission mode, the power allocation problem in full-duplex relay transmission mode is non-trivial. Thus, we provide a novel hierarchical dual method to solve the power allocation problem in full-duplex relay transmission mode. In addition, in half-duplex relay cooperative transmission mode, the joint transmission of both source and relay is taken into account, and we provide a simple and insightful power allocation scheme. Results show that the system throughput enhances significantly compared to previous works. Thirdly, we investigate a max-min weighted SINR problem in a full-duplex multi-user MIMO system, where a full-duplex-capable base station equipped with multiple antennas communicates with multiple half-duplex downlink and uplink users under the same system resources. Instead of optimizing the joint uplink and downlink max-min weighted SINR, we consider a more practical scenario where the downlink minimum weighted SINR is maximized under specific SINR constraints for uplink users. Moreover, the optimization is conducted by jointly considering the base station transmit power, uplink transmit power, and base station transmit and receive beamforming. This optimization problem is therefore subject to multiple uplink SINR constraints and multiple transmit power constraints. Due to the SINR constraints, negative matrix components arise and hence the optimization problem cannot be solved by the Perron-Frobenius theory directly. With fixed base station transmit and receive beamforming, we first optimize the max-min weighted SINR problem under multiple uplink SINR constraints and a single power constraint, and show how the subgradient projection-based method can be applied to optimize the problem under multiple-power-constraint conditions. Then we derive the network duality of the same problem, i.e., fixed base station transmit/receive beamforming with multiple uplink SINR constraints and a single power constraint. To solve the original problem, we propose an optimization algorithm that iteratively updates (i) the transmit power vector and receive beamforming in the primal domain, and (ii) the dual transmit power vector and transmit beamforming in the dual domain. Moreover, the algorithm, which is also based on the subgradient projection-based method, is proven to converge under appropriate initialization parameters. With network duality, we avoid optimizing coupled transmit beamforming in the primal domain and instead are able to optimize individual transmit beamformers easily in the dual domain. Simulation results show that our proposed algorithm has a fast convergence rate and leads to a better performance compared to other optimization techniques that do not jointly considered all parameters. Finally, energy efficiency of a full-duplex relay system under the total power constraint and fixed circuitry power consumption is studied. An optimization problem is formulated towards maximizing the system energy efficiency. Unfortunately, this problem is non-trivial and cannot be solved by conventional fractional programming methods, such as the Dinbelbach's method. To resolve this issue, an algorithm called sequential parametric convex approximation-Dinbelbach is proposed. Simulation results show that the proposed algorithm can converge to the global optimum very quickly.