Precoder Design for MIMO Systems with User Cooperation and Non-Orthogonal Multiple Access


Student thesis: Doctoral Thesis

View graph of relations


Related Research Unit(s)


Awarding Institution
Award date4 Sep 2020


The ever-increasing demand of high system capacity, reliable quality of service (QoS), and low processing latency has been calling for new techniques in wireless communications, especially in the upcoming fifth-generation era, to support massive connectivity with stringent QoS requirement.

Among existing promising candidates for performance enhancement, multiple-input multiple-output (MIMO) with precoding technique has been widely used to provide diversity and/or multiplexing gains via degrees of freedom in the spatial domain. Cooperation among mobile users over device-to-device channels can be exploited to improve the QoS of users who have poor channel gains to the base station (BS). The system capacity can be enhanced by power-domain non-orthogonal multiple access (NOMA) which allows multiple users to transmit data simultaneously. With the use of superposition coding and successive interference cancellation (SIC), NOMA has superior spectral efficiency over the conventional orthogonal multiple access.

This thesis focuses on transmit-receive strategies and precoder designs in the uplink of MIMO networks where user cooperation and NOMA protocol are studied.

In the first study on user cooperation, idle users act as relay nodes to help an active user (source node) with poor channel gain to transmit to the BS (destination node). The works focus on the source and relay precoder design of a generic two-hop amplify-and-forward (AF) MIMO relay network with one source, multiple relays, and one destination for general scenarios. A robust precoder design based on statistical channel state information (CSI) for MIMO-AF multiple-relay networks in double correlated Rician fading channels is developed, where all the nodes are equipped with multiple antennas and the source-to-destination direct link is included for a general formulation. A closed-form upper bound of average capacity is derived and adopted as the criterion for designing suboptimal precoders. Based on the capacity bound, the source precoder, relay beamforming matrices, and relay power allocation can be optimized iteratively by the proposed design procedure to give substantial capacity enhancement with fast convergence, good numerical stability, and efficient computation. The proposed iterative algorithm with trivial initialization exhibits almost the same results as the near optimum for both perfect CSI and statistical CSI. The proposed source precoder design and the new relay power adaption method outperform upon the augmented Lagrangian method with respect to the numerical stability, computation time, and convergence rate. The generality of the proposed scheme makes the design procedure applicable to MIMO-AF relay networks with/without the direct link under either perfect CSI or statistical CSI for different simplified Kronecker channel models.

In the second study on NOMA protocol, multiple active users transmit data simultaneously. A novel MIMO-NOMA scheme with multi-group detection is proposed and different precoder designs are developed. In the proposed scheme, users are divided into groups for detection at the BS. The inter-group interference is cancelled by the SIC, while the intra-group interference is mitigated by transceivers. The study consists of two parts: (i) In Part 1, assuming error-free SIC, system capacity is adopted as the design criterion. The BS can adopt zero-forcing (ZF) or QR-based equalizers in the detection of each group. For using the ZF detection, an approximate signal-to-interference-plus-noise ratio (SINR) is derived for each user. An iterative power and beamforming update procedure is developed for obtaining the user precoders. In particular, the beamforming update is cast as the maximization of the product of Rayleigh quotients. For this generic problem, three suboptimal algorithms are proposed, namely, a minorization-maximization based iterative algorithm, and two vector-basis-based non-iterative algorithms. The three algorithms are shown to give performance very close to the best out of multiple local optima obtained by the conventional gradient method with steepest descent line-search. The precoder design can effectively mitigate the intra-group-inter-user correlation and thus helps maintain the accuracy of the adopted approximate SINR. For using the QR-based detection, we derive the analytical user SINRs by following the Gram-Schmidt process and then develop a sequential precoder design. Simulations show the effectiveness of the proposed ZF- and QR-based multi-group MIMO-NOMA schemes, both of which outperform upon existing signal alignment (SA) based MIMO-NOMA schemes in terms of total transmit power for various system configurations. By utilizing the intra-group SIC on top of the inter-group SIC, the QR-based multi-group MIMO-NOMA scheme can save transmit power as compared to the ZF-based scheme, and the power saving is increased as the group size enlarges.

(ii) In Part 2, the precoder design is extended to imperfect SIC scenarios. For the proposed MIMO-NOMA scheme with multi-group detection, we consider practical modulation and symbol demodulation for the investigation of the SIC residuals. Performance analysis of the system in the presence of the SIC residuals is conducted. By deriving an approximate SIC-residual-power coefficient, an approximate SINR is obtained. In doing so, the effect of the SIC residuals is explicitly addressed in the precoder design for minimizing the total transmit power subject to users’ maximum symbol error probability (SEP) constraints. A suboptimal precoder design algorithm with efficient “logarithmic” golden-section search is developed. Simulation results show the effectiveness of the derived design criterion and the developed design algorithm, and also validate the detection error analysis. By explicitly considering the mutual impact between the instantaneous SINR and SIC error, the proposed SIC residual formulation always provides precoder design solutions that satisfy pre-specified SEP requirements. This avoids the situation that the conventional fixed-SIC-error-fraction model suffers from infeasibility issues under certain conditions such as large fractional-SIC-factor, large number of users, and small number of BS antennas.