The Papr Reduction For Large-Scale MIMO-OFDM Downlink Systems

In this thesis, we consider the problem of peak-to-average power ratio(PAPR) reduction in orthogonal frequency-division multiplexing(OFDM) based massive multipleinput multiple-output(MIMO) downlink systems. In the future massive MIMO-OFDM systems, the base station(BS) will be equipped with a large number of antennas and serving a much small number of users simultaneously. The large-scale transmit antennas provide excess degrees-of-freedom(DoFs), which enable us to develop efficient PAPR reduction schemes without distinct performance loss. In this paper, we have developed two PAPR reduction schemes for massive MIMO-OFDM systems by the use of DoFs at the transmitter.We first discuss the basic idea of PAPR reduction using the DoFs of precoding,and then develop an efficient Bayesian PAPR reduction method, in which the multi-user(MU) precoding constraint and OFDM modulation are jointly formulated as an underdetermined linear equations, and a Bayesian method is used to search from the underdetermined system for the low-PAPR solution. Specifically, we propose a hierarchical truncated Gaussian mixture prior to model the characteristic of low-PAPR solution, and then develop an efficient Bayesian inference method to estimate the low-PAPR signal utilizing the variational expectation maximization(EM) and generalized approximate message passing(GAMP) algorithm. Simulation results show our method achieves superior PAPR reduction and fast convergence speed at Nyquist sampling rate.Then, we analyse comprehensively the drawbacks of the previous PAPR reduction method, and propose a novel PAPR reduction approach by the use of additive artificial disturbance. Specifically, we introduce an artificial disturbance into the frequencydomain signal, which is able to reduce the PAPR of the time-domain signal, and meanwhile will not generate in-band distortion and out-of-band radiation. We formulate the above PAPR reduction approach as a convex optimization problem, and then develop an efficient optimization algorithm to search for an approximate solution making use of the∞-norm proximal operator and alterative direction method of multipliers(ADMM) algorithm. Instead of optimizing the high-dimensional oversampled time-domain signal, in the proposed method the frequency-domain signal is optimized. Therefore, it is very robust to oversampling. Furthermore, the proposed scheme is independent with precoding,thus it is compatible with existing real systems. Simulation results show that, in the case of oversampling, the proposed PAPR method is able to explore the DoFs at the transmitter more efficiently than the methods based on precoding, and can achieve a substantial improvement in terms of both the PAPR reduction and convergence speed.