毫米波超密集网络中高能效的多点协作波束赋型技术研究

Mmwave ultra-dense networks are able to satisfy the capacity and coverage requirements of the next generation 5G systems. However, the energy consumption of such networks will increase dramatically. In order to deal with this problem, coordinated multi-point beamforming (CMBF) has been proposed in the literatures. The research of energy-efficient CMBF in the Mmwave ultra-dense networks has following challenges: the asymmetry of the type and number of cooperative base stations, the limitation of minimum Qos guarantee and single target optimization, and the channel error caused by backhaul will affect the networks performance seriously. In this proposal, the energy-efficient CMBF with asymmetric conditions, the requirements of new business, and robust is explored. This proposal mainly focuses on the following points, 1) the theoretical model of coordinated multi-point and energy-efficiency forming of cooperative cluster considering base station dormancy will be proposed; 2) energy-efficient CMBF algorithm under asymmetric conditions will be designed considering the impact and allocating users to exploit the degree of freedom; 3) energy-efficient CMBF algorithm to ensure quality of experience (QoE) through multi-objective optimization will be given; 4) a robust energy-efficient CMBF algorithm will be proposed, in which the CSI error distribution will be taking into account. The research achievements are expected to promote the application of CMBF technology of Mmwave ultra-dense networks and provide new solutions for energy efficiency improvement.

毫米波超密集网络可以满足未来5G系统对网络容量和覆盖的要求，但同时也带来了网络能耗剧增的问题，多点协作波束赋型（CMBF）技术是解决该问题的有效方法。毫米波超密集网络中的CMBF能效研究主要面临以下难题：协作基站的种类和数量出现不对称、Qos最低保证及单目标优化的局限性、回传链路限制导致的信道误差严重影响网络性能。针对上述难题，本项目研究非对称条件、满足新业务需求、鲁棒的高能效CMBF技术，具体内容：1）构建多点协作理论模型，提出考虑基站休眠的高能效协作簇的划分方法;2）考虑非对称的影响并通过分配用户提高空间资源利用率，设计非对称条件下的高能效CMBF算法;3）通过多目标优化设计保证用户体验质量（QoE）的高能效CMBF算法；4）利用CSI误差分布特性设计鲁棒的高能效CMBF算法。本项目的研究成果对于推动毫米波超密集网络的CMBF技术应用具有重要的意义，为网络能耗问题提供新的解决方案。

毫米波、大规模MIMO系统和超密集网络可以有效提高网络容量和覆盖，但也带来了诸多挑战和难题。其中，网络能耗剧增的问题尤为突出，高效的波束赋型技术是解决该问题的有效方法。本项目主要针对系统波束赋型设计的优化目标单一、回传链路假设的局限性、信道估计误差影响大等问题进行研究。通过分别利用模拟退火分析、迭代优化方法和深度学习方法等优化了系统的波束赋型；针对量化与CSI偏差实现了波束赋型优化策略；基于凸优化和深度学习理论开发并实现了预编码设计新算法。分别从单小区、多小区等多种情况进行系统部署，构建考虑能效优化结合超密集网络的理论模型，基于凸优化分析系统性能，取得了很好的部署效果。以毫米波大规模MIMO系统的信道新特点为关键出发点，结合5G网络的拓扑特性，研究基于高性能的信道估计优化策略，实现神经网络于与系统特性的最佳匹配，提高信道估计的精度和准确性，为后续的预编码技术提供有力保证。基于毫米波与大规模MIMO系统特点，同时考虑不同系统资源限制和优化目标，分别实现基于退火、迭代凸优化、深度学习的预编码设计，保证系统性能。本项目的研究成果对于推动毫米波、大规模MIMO、超密集网络的波束赋型技术应用具有重要的意义，为网络设计和网络能耗问题提供新的解决方案。