EAGER: Hybrid Precoding for Massive MIMO Communication Networks

EAGER：用于大规模多输入多输出毫米波无线通信网络的混合预编码由于智能手机、自动驾驶汽车和移动设备的日益普及，未来的智能和互联社区将对无线通信提出巨大需求。大规模多输入多输出（MIMO）技术与毫米波（mmWave）频谱利用相结合，被认为是实现下一代无线网络数据速率大幅提高的关键突破。在大规模MIMO系统中，在基站上使用大量天线来同时与许多移动用户通信。然而，如果采用每个天线需要一个射频（RF）链的传统方法，那么大量的天线可能导致过高的成本和功耗。本课题的重点是研究一种新的混合预编码设计，它不仅可以减少射频链的数量，而且可以最大限度地提高数据速率。混合预编码由模拟预编码器和数字预编码器组成，其中数字预编码器由少量射频链实现，模拟预编码器由移相器实现。因此，可以大大降低大规模MIMO系统的成本、复杂性和功耗。所提出的研究将对下一代无线网络的设计和开发产生重大影响，其中大规模MIMO和毫米波是关键的使能技术。研究结果将透过设计新的课程计画，整合到电子工程与电脑工程专业的课程中。研究结果将通过会议发言和期刊出版物广泛传播。此外，该项目将有助于增加代表性不足的少数民族的参与，并加强外联活动，以吸引女学生从事工程职业。该项目旨在研究混合预编码设计方法，该方法可以大幅降低成本，复杂性和功耗，同时接近全连接，无约束的大规模MIMO系统的最佳性能。本课题将多用户大规模MIMO系统的混合预编码设计表述为模拟和数字预编码的联合优化，动态资源分配包括子阵列选择、功率分配和调制编码速率选择。联合优化将使混合系统显著减少射频链的数量，以较小的成本实现与全连接大规模MIMO系统相似的性能。动态资源分配将有助于在给定信道条件下实现最佳吞吐量。此外，该方法利用有限字母输入和统计信道状态信息（CSI）代替理想的高斯输入和瞬时CSI，从而提高了优化后的预编码器对实际系统的鲁棒性。该项目的目标预计将通过三项具体任务来实现。首先，可实现的数据速率的理论研究将解决混合预编码的性能和成本之间的基本权衡。其次，算法研究将推导出求解NP-hard优化问题的低复杂度解。第三，将应用机器学习技术来学习控制在线资源分配和混合预编码的马尔可夫决策过程的特征和转移概率。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

EAGER: Hybrid Precoding for Massive Multiple-Input Multiple-Output Millimeter Wave Wireless Communication NetworksFuture smart and connected communities will place tremendous demand on wireless communications due to the ever-growing popularity of smartphones, autonomous vehicles, and mobile devices. The massive Multiple-Input Multiple-Output (MIMO) technology, in combination with millimeter wave (mmWave) spectrum utilization, is considered as a key breakthrough for enabling enormous data-rate increase for next generation wireless networks. In massive MIMO systems, a very large number of antennas is employed at the base station to communicate many mobile users simultaneously. However, this large number of antennas can lead to prohibitive cost and power consumption if conventional approach which requires one radio frequency (RF) chain per antenna is adopted. This project focuses on novel hybrid precoding designs which will not only reduce the number of RF chains but also maximize the data rate. The hybrid precoding consists of analog and digital precoders, where the digital precoder is realized by a small amount of RF chains, and the analog precoder is realized by phase shifters. Therefore, the cost, complexity and power consumption of massive MIMO systems can be reduced dramatically. The proposed research can have a large impact on the design and development of future generation of wireless networks, for which massive MIMO and mmWave are key enabling technologies. The research results will be integrated into the classes for electrical engineering and computer engineering majors through designing new course projects. Research findings will be broadly disseminated through conference presentations and journal publications. Moreover, the project will help increase participation of under-represented minorities and enhance outreach activities to attract female students to careers in engineering. This project aims to investigate hybrid precoding design methods that can drastically reduce the cost, complexity and power consumption while approaching the optimal performance of fully-connected, unconstrained massive MIMO systems. The project formulates the hybrid precoding design of multi-user massive MIMO systems into a joint optimization of analog and digital precoders with dynamic resource allocation which includes subarray selection, power allocation, and modulation-coding-rate selection. The joint optimization will enable the hybrid system with a significantly reduced number of RF chains to achieve similar performance of fully-connected massive MIMO systems at a fractional cost. The dynamic resource allocation will help to achieve best throughput for given channel conditions. Furthermore, the proposed approach utilizes finite-alphabet inputs and statistical channel state information (CSI) instead of the idealistic Gaussian inputs and instantaneous CSI, thus improving the robustness of the optimized precoders for practical systems. The objective of this project is expected to be accomplished by three specific tasks. First, theoretical studies of the achievable data rates will address the fundamental tradeoff between performance and cost of hybrid precoding. Second, algorithm research will derive low-complexity solutions to solve the NP-hard optimization problems. Third, machine learning techniques will be applied to learn the features and transition probabilities of the Markov decision processes governing the online resource allocation and hybrid precoding.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.