Abstract
In this thesis, we propose a downlink multiple-input multiple-output non-orthogonal multiple access (MIMO-NOMA) system. We first consider a MIMO-NOMA system with a base station and four users, where the four users are divided into two clusters and the inter-cluster interference is induced in the downlink transmission. With the fact that the eigenvalues of the channel matrix would significantly affect the capacity of a MIMO system, we then propose a joint clustering and precoding algorithm for the MIMO-NOMA system such that not only the inter-cluster interference can be effectively eliminated but also the sum capacity can be enhanced. Specifically, a set of precoder pairs (one pair having two precoders for the two clusters) are first constructed based on the eigenspaces of channel matrices, and then one of them is selected to maximize the sum capacity. Compared to a MIMO orthogonal multiple access (OMA) system, the proposed joint algorithm can provide a larger capacity gain, but involve a high complexity issue due to the need of exhaustive search for the best precoder pair. To reduce the search complexity, we further explore the channel gain of each user and construct a reduced-size precoding set for selection. The modified clustering and precoding algorithm can achieve a close sum capacity to the original one, but reduce the search complexity by a factor of about three. Combined with orthogonal frequency-division multiplexing (OFDM), the proposed MIMO-NOMA scheme can also be extended to the general multiuser case, where every four users share a subband for MIMO-NOMA transmission.