EAGER: A Novel Approach to Achieve Real-time Wireless Network Optimization

资源优化是无线网络研究中的一个基本问题。随着无线技术变得越来越复杂，资源优化问题也变得更加复杂。例如，当前和未来蜂窝网络设计中的一个主要技术挑战是解决涉及频谱和时间域中的大量资源单元的调度问题。虽然已经提出了一些调度器来实现最优(或接近最优)目标，但由于计算时间过长，它们通常不能在现场使用。取而代之的是，运营商不得不采用简单的调度器，并满足于远远不是最优的性能。这个渴望项目的目标是通过探索一种新的方法来解决这个根本问题，从而使实时资源优化成为可能。所谓实时，我们的意思是，资源优化问题的解决方案可以在现场容易使用的时间分辨率中找到。我们的方法是将一个复杂的资源优化问题分解成大量极其简单的子问题，这些子问题可以精确地匹配到给定的图形处理单元(GPU)计算平台上。通过独立求解各个子问题，同时并行求解各个子问题，可以满足实时调度对时间的严格要求。拟议的想法如果成功，将有可能彻底改变在该领域进行资源优化的方式。拟议的研究活动将对制订新的教育课程和扩大任职人数不足群体对计算机的参与产生深远影响。将根据这项研究开发新的教材，并将在弗吉尼亚理工大学和其他大学的课堂上使用。将通过我们的Wireless@VT教育和研究计划，特别努力扩大女性和代表性不足的学生对计算机的参与。最后，两位研究人员将继续努力，在网络会议上支持N2女性活动，以培养女性初级研究人员成为网络社区的领导者。我们必须仔细解决一些重大的技术挑战，然后才能从我们提出的想法中获益。在这个项目中，研究人员计划探索以下两个主要领域作为概念验证：(I)如何将资源优化问题以最佳方式分解并映射到给定的最先进的GPU平台，以便能够实时获得解决方案；(Ii)如何将该方法应用于下一代蜂窝网络以及其他(非蜂窝)无线网络的实时优化调度。提出的方法是跨学科的，涉及从计算(GPU)和优化(分解)领域探索新的方法和工具来解决无线网络中的基本问题。所提出的思想将在现成的低成本GPU平台上进行验证。一旦建议的想法被证明是有效的，我们预计将看到一种新的实用方法，以实现实时资源优化和在现实世界中显著的绩效改进。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

Resource optimization is a fundamental problem in wireless network research. As wireless technologies become increasingly sophisticated, resource optimization problems also become more complex. For example, a major technical challenge in the design of current and future cellular networks is to solve scheduling problems involving a large number of resource units in spectral and temporal domains. Although some schedulers have been proposed to achieve optimal (or near-optimal) objectives, they typically cannot be used in the field due to excessive computation time. Instead, operators have to adopt simple schedulers and settle for performance far from optimal. This goal of this EAGER project is to tackle this fundamental problem by exploring a novel approach that could make real-time resource optimization possible. By real time, we mean that a solution to the resource optimization problem can be found in a time resolution that can be readily used in the field. Our approach is to decompose a complex resource optimization problem into a large number of extremely simple subproblems that can be matched precisely into a given Graphical Processing Unit (GPU) computing platform. By solving each small subproblem independently and all the sub-problems in parallel, it is possible to meet the stringent timing requirement for real-time scheduling. The proposed idea, if successful, has the potential to revolutionize how resource optimization is performed in the field. The proposed research activities will have a profound impact on developing new educational curricula and broadening participation in computing from underrepresented groups. New teaching materials will be developed from this research and will be used in classrooms at Virginia Tech and other universities. Special efforts to broaden participation in computing by female and underrepresented students will be made through our Wireless@VT education and research programs. Finally, the two investigators will continue their efforts to support N2 Women events at a networking conference to nurture female junior researchers to become leaders in the networking community. Some major technical challenges must be carefully addressed before we can reap the benefits of our proposed idea. In this project, the investigators plan to explore the following two thrust areas as proof-of-concept: (i) how to optimally decompose and map a resource optimization problem into a given state-of-the-art GPU platform so that the solution can be obtained in real time; (ii) how to apply the methodology for real-time optimal scheduling for next generation cellular networks as well as other (non-cellular) wireless networks. The proposed approach is interdisciplinary and involves exploring new methods and tools from the fields of computing (GPU) and optimization (decomposition) to address fundamental problems in wireless networks. Validation of the proposed ideas will be carried out on off-the-shelf low-cost GPU platform. Once the proposed idea is proven to be valid, we expect to see a new and practical methodology to achieve real-time resource optimization and significant performance improvement in the real world.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.