User Pairing for Mobile Communication Systems with OSC and SC-FDMA Transmission
Michael A. Ruder · OPUS FAU (Kooperativer Bibliotheksverbund Berlin-Brandenburg (KOBV), on behalf of the Universitätsbibliothek Erlangen-Nürnberg) · 2014
The main subject of this thesis is user pairing for mobile communication systems. User pairing in general describes a user selection process, where we choose the users that should transmit on the same time and frequency resource in the same cell. Typically, user pairing is optimized for a prescribed performance criterion, like bit error rate (BER) or achievable data rate. The spectral efficiency of a communication system employing user pairing is thereby increased compared to a system without coordinated pairing. In this thesis, user pairing is considered for two systems and the spectral efficiency gains compared to randomly chosen pairs are shown to be impressive. The first part of this thesis focuses on downlink transmission with orthogonal sub-channels (OSC). The evolution of the Global System for Mobile Communications (GSM) led to the standardization of Voice services over Adaptive Multi-user channels on One Slot (VAMOS). The aim of VAMOS is to double the spectral efficiency of GSM voice transmission by deliberately transmitting two Gaussian minimum-shift keying (GMSK) signals on the same time and frequency resource in the same cell, while guaranteeing backward compatibility for legacy receivers. The phase of the signal of the second user is rotated by 90° compared to that of the first user, which is referred to as OSC transmission. In this work, downlink OSC transmission is considered, where different transmit powers can be assigned to the two users of an OSC pair. Advanced receiver architectures are necessary to improve the separability of the users of one pair. Therefore, algorithms for channel estimation, equalization, and interference cancellation for an OSC downlink transmission are investigated in this thesis. Two novel algorithms for joint estimation of the channel and the power imbalance between the users, i.e., the subchannel power imbalance ratio (SCPIR), are proposed. The Cramer-Rao lower bound (CRB) w.r.t. the mean-squared error (MSE) of the joint estimation is derived for the first time for this estimation scenario. A comparison of the MSE of the suggested algorithms with the lower bound given by the CRB exhibits their excellent performance. Moreover, several equalization and interference cancellation algorithms are developed and evaluated via simulations. Furthermore, a novel asynchronous co-channel interference cancellation technique, based on modeling the equalizer metric by a Generalized Gaussian probability density function (pdf), is proposed. Additionally, radio resource allocation (RRA) performed by the base station (BS) is considered for OSC transmission. Power allocation as well as user pairing, i.e., the decision which pairs transmit in the same time slot and frequency resource, have to be optimized with the aim to maximize the network capacity. A practical RRA algorithm for OSC transmission is proposed and evaluated by network simulations for a GSM VAMOS network. Simulation results reveal significant frame error rate (FER) performance gains for the proposed downlink OSC receivers compared to state-of-the-art receiver architectures. Furthermore, simulations of a GSM VAMOS network employing the proposed RRA algorithm jointly with the novel receiver algorithms exhibit a network capacity gain of about 100 % compared to non-OSC transmission. The second part of this thesis considers the uplink of Long Term Evolution (LTE). In contrast to the downlink of LTE Release 8, where orthogonal frequency-division multiple access (OFDMA) and single user multiple-input multiple-output (MIMO) transmission are employed, the first release of LTE specifies only a single transmit antenna single-carrier frequency-division multiple access (SC-FDMA) transmission for the uplink. To enhance the spectral efficiency in the uplink, a virtual MIMO (V-MIMO) transmission can be employed, where multiple users transmit on the same time and frequency resource. Here, multiple receive antennas at the BS facilitate a separation of the signals of a user pair. By employing the proposed V-MIMO transmission, the spectral efficiency is impressively improved compared to a single-input multiple-output (SIMO) transmission. Various receiver algorithms for a V-MIMO SC-FDMA transmission are proposed in this thesis to separate the signals of the users of each pair. An algorithm for reference signal based channel interpolation and prediction, which is necessary to obtain channel state information (CSI) for RRA, is presented and an MSE expression for the channel acquisition error is derived. Furthermore, power allocation and beamforming (BF) with quality of service (QoS) requirements are studied for a V-MIMO SC-FDMA transmission and closed form solutions for zero-forcing (ZF) equalization at the receiver are obtained. Simulation results for random user pairs exhibit significant transmit power savings for the BF algorithm compared to constant power allocation for all subcarriers. Finally, several joint user pairing/grouping and RRA algorithms for SC-FDMA V-MIMO transmission are proposed and studied. First, different criteria for user grouping are introduced. Then user pairing in time and frequency direction are considered. For the latter, we investigate joint user grouping and frequency allocation. The combinatorial optimization problems for the considered scenarios are stated and novel suboptimal algorithms with reduced complexity are proposed. Furthermore, user pairing with QoS constraints, minimizing the required sum transmit power with or without BF, is investigated. A study of the influence of inaccurate CSI on the user pairing performance concludes the thesis. All proposed algorithms are evaluated via simulation of a V-MIMO SC-FDMA transmission. It is shown that significant performance gains are achieved compared to random user pairing. An exhaustive search would be needed to find the optimal solution to the joint user pairing and frequency allocation problem. However, the proposed suboptimal algorithms provide a close-to-optimum performance as well as a low computational complexity.