Key Technologies For Massive MIMO Beamforming System

The massive multiple input multiple output(massive MIMO)beamforming technology is one of the key physical-layer technologies in the 5th generation(5G)mobile communication system and the fundamental technologies for future wireless communication system.As an important method to greatly improve the performance of physical-layer for wireless communication system,the massive MIMO beamforming technology is widely acknowledged to be the key academic research area for the study of information and communication engineering,as well as the heated research focus for the wireless industry.With the emerging of the standardization process for 5G and the fast development of wireless communication theory,the massive MIMO beamforming research has become a widely discussed topic with great research value.This work is conducted concerning the technology and scenario of massive MIMO beamforming.First,for the wireless channel modeling work,considering the specialty of wireless channel in 5G physical-layer,since the conventional model for wireless channel cannot precisely describe the characteristics of massive MIMO channel,the innovative channel model for this scenario should be established for the research of beamforming.Also,although the wireless multicast beamforming technology can achieve multiple magnitude of performance improvement for spectrum efficiency and energy efficiency comparing to traditional beamforming technologies,there has been little research exploration for this technology under the scenario of 5G.To address these issues,in this work the mutual coupling channel model is introduced for massive MIMO channel modeling,along with the hybrid beamforming strategy utilizing both wireless multicast and linear precoding.The signal model as well as the optimization beamforming strategy are derived with asymptotic analysis under massive MIMO scenario.To address the multi-user interference issue,the null-space based method is introduced to pre-process the beamforming signal.The simulation system for the proposed hybrid beamforming transmission system is developed,based on which the comparisons between different beamforming strategies and channel models are provided and analyzed.The main contributions include utilizing the mutual coupling model based massive MIMO channel model and the combination of null-space interference cancelling technology to derive and analyze the hybrid beamforming transmission system,providing the foundation of system model for further researches.Furthermore,considering the impact of imperfect channel information issue which affects the performance of precoding in massive MIMO beamforming system,to describe the physical-layer user data transmission scenario,the large-scale Markov decision process(MDP)and reinforcement learning(RL)based system model is developed which imcorporates the conventional convex-optimization based methods and aims at modeling the beamforming strategy for massive MIMO system.The abstract scheduling optimization problem under the average reward criterion is proposed,based on which the equivalency between the MDP based beamforming scheduling problem and the original problem as well as the existance of optimal solution are explored and analyzed.The equivalent scheduling problem is proposed for the wireless multicast beamforming technology scenario.Utilizing the asymptotic characteristics of massive MIMO,the proposed problem is decomposed into a series of RL subproblems,establishing the partially observed scheduling subproblems with policy estimation method.Last,the policy gradient theorem based RL approach is adopted to derive the algorithm to solve the scheduling problem.Based on the previous theory,the methodology is extended to the physical-layer security scenario in massive MIMO beamforming system.The secrecy capacity is regarded as the optimization target,combining the artificial noise shaping and beamforming to derive the signal model the optimization problem.Based on the previous RL based analytical method,the scheduling problem is established with asymptotic analysis in massive MIMO scenario.The physical-layer security subproblem is derived and analyzed using the RL architecture.Simulation results provide the comparisons between the RL based optimal beamforming policy and the randomized beamforming policy with the channel information being unobserved,which proves the rationality of the proposed method.The main contributions include extending the RL and MDP based scheduling technology to the massive MIMO beamforming system and deriving the model-free RL based beamforming optimization algorithm,which does not suffer from the imperfect channel information issue and can be regarded as supplementary method for the traditional convex-optimization based signal processing method.In the last,considering the systematic level of massive MIMO beamforming technology,this work also focuses on the user scheduling problem in multi-cell scenario.The signal model and mobility model are derived based on the massive MIMO beamforming technology with the asymptotic analysis.The high-dimensional symmetrical large-scale MDP based model is established to describe the user scheduling problem.Considering the cell reselection scenario for heterogeneous system,the MDP based user scheduling problem is further proposed.In order to decompose the large-scale MDP problem,the multi-agent RL(MARL)theory is proposed with the definitions and conditions for three types of symmetrical property.For the proposed strong symmetrical MARL problem,the MARL subproblem policy is utilized for policy estimation with theoretical analysis for the policy estimation error.Applying the proposed theory in the previous problems,the MARL based user scheduling problem is proposed.The model for solving the MARL subproblem based user scheduling problem is established with algorithmic solution.Simulation results are provided for the user scheduling problems in different scenarios along with the policy estimiation error for MARL problems.The main contributions include establishing multiple user scheduling model for massive MIMO beamforming system,proposing the symmetrical MARL theory and its policy estimation error theory,proposing the symmetrical MARL based user scheduling model and the simulation results for the proposed methods.