Beam Management And Resource Scheduling Techniques Of Millimeter Wave System

Millimeter-wave(mm Wave)communication technology is one of the critical technologies in 5th and future 6th generation mobile communications.It can use large bandwidth to meet the needs of users for high-rate,low-latency communications.Nevertheless,due to the high-frequency band characteristics,mm Wave communications suffer from disadvantages such as excessive path fading and restricted coverage.To compensate for the impact of path fading on communication quality,beamforming gain is obtained by beam management technology.Beam management is challenging to perform quickly in a wide search space.Therefore,accurately completing and maintaining beam alignment with low limited overhead is one of the keys to mm Wave communication.However,when the mm Wave communication suffers severe path loss,the stability of the link cannot be maintained even with precise beam alignment.Thus,the mm Wave communication system integrating the sub-6GHz band has been presented to provide high data rates while ensuring reliable coverage.However,the integrated communication system presents new challenges for resource scheduling and precoding.On the one hand,due to differences in signal propagation characteristics and bandwidth availability,the reachable rates in the mm Wave and sub-6GHz bands are significantly different,which poses challenges to the scheduling of combined resources.On the other hand,improper resource scheduling can lead to increased interference between users,which affects the performance of precoding.To address the above problems,the research in this thesis mainly includes mm Wave beam training in the initial phase,beam tracking in the maintenance phase,and joint optimization of dual-frequency resource scheduling and precoding,as summarized below:To decrease the overhead of mm Wave beam training in the establishment or recovery stage of the communication link,this thesis investigates the feature fusion process of mm Wave channel power leakage and sub-6GHz channel state information.The beam training scheme based on a multimodal fusion network is designed.Compared with the uni-modal counterparts(mm Wave only or sub-6GHz only),the complementarity and interactivity between different modalities make the proposed network more accurate and robust.Specifically,we propose a multimodal fusion network that utilizes the instantaneous signal received from the mm Wave wide beam and sub-6GHz channel state information(CSI)to estimate the optimal narrow beam.Furthermore,the additional mm Wave narrow beam decided by the predicted probabilities is employed to calibrate beam direction further.Moreover,we introduce two optimization strategies based on the proposed multimodal fusion network.Firstly,to improve the robustness to noise in wide beam training,a long and short memory(LSTM)network is employed to track the motion direction of the user and calibrate the mm Wave beam.The wide beam for training is selected according to the previous prediction probabilities to reduce the training overhead.Secondly,given that beam training focuses on partially specific information,to focus on its related features,we introduce the self-attention mechanism to improve the global consistency of the network.The proposed scheme can achieve significantly more accurate results with less overhead than traditional and existing deep learning-based methods.To solve the problem of low tracking accuracy for beam tracking in the communication link maintenance stage,a model-driven deep learning network is designed for beam tracking,which combines traditional signal processing methods with convolutional neural networks.Specifically,traditional signal processing is designed to drive the network to enhance the feature extraction process.In contrast to previous work based on deep learning,due to model-driven,fewer pilot sequences are required to track the arrival angle of the main path.Due to the limitation of manufacturing cost and the influence of environment,mm Wave antennas often have non-ideal characteristics.To ensure the accuracy of beam tracking,periodic diagnostics of mm Wave antennas are required.However,offline antenna diagnostics are not feasible to minimize the impact on the efficiency of the communication system.In addition,the incoming wave direction estimation and the non-ideal estimation are coupled problems.We propose an adaptive iterative diagnostic algorithm based on clustering block sparse Bayesian learning(CBSBL)to address the coupling problem.The sparse signal is constructed by approximating the difference between the radiation pattern of the fault-free reference antenna and the non-ideal antenna.The structural clustering method is used to address the sparsity characteristics of the blocked antennas and the correlation characteristics of adjacent coefficients.The estimation accuracy is further enhanced by encouraging the dependence between the adjacent coefficients by controlling the neighboring hyperparameters.To solve the joint optimization problem of resource scheduling and precoding,a novel resource scheduling and precoding algorithm based on collaborative sensing is investigated in the mm Wave systems integrating the sub-6GHz band.We formulated the collaborative optimization problem of resource scheduling and precoding,aiming to maximize the sum of data rates while meeting minimum acceptable rate requirements and ensuring proportional fairness.Considering the spatial similarity between the mm Wave channel and sub-6GHz channel,the prior information of the mm Wave channel predicted by the sub-6GHz band is generated to process sensing weight,which realizes fast dual-mode network resource scheduling.Furthermore,a low-complexity two-stage precoding scheme for dual-mode networks is proposed based on sub-6GHz spatial information.Low-complexity analog precoding is realized in the first stage based on sub-6GHz spatial information.In the second stage,inspired by the correlation between minimum mean square error(MMSE)and weighted sum-rate(WSR),we proposes an alternating optimization algorithm to find a local WSR optimum with low complexity.Through extensive simulations and comparisons with heuristic algorithms,the results demonstrate the apparent advantages of the proposed collaborative sensing optimization algorithm.Most importantly,we can achieve a performance that approaches an upper bound.