SWIFT: Instantaneous Feedback-based Adaptive Communications and Networks

本研究计划探讨下一代无线通讯系统的即时回馈自适应通讯（IFAC）。所提出的IFAC使无线通信信号的亚比特级实时适应能够利用基于接收器瞬时反馈的频谱效率、可靠性和延迟的动态权衡。大多数传统系统都是基于平均噪声功率设计的消息级链路自适应，而没有利用子比特级噪声观测。这种消息级反馈方案对于需要在大规模机器类型通信网络中进行超可靠低延迟通信（URLLC）的新兴和未来应用程序来说是次优的，因为这些网络中的设备数量可能远远大于当今的消费者移动蜂窝网络。该研究计划将研究基于瞬时反馈的无线通信系统的可行性，以获得更接近理论信道容量和可靠性极限的优越性能，优于最先进的传统方法。提议的IFAC作为下一代（5G以上）URLLC的关键使能技术具有巨大潜力，其目标是100微秒延迟和99.99999%的短消息传输可靠性。通过IFAC实现如此严格的URLLC目标可以实现各种类型的新应用，包括自主微型机器人，智能自动化工厂和残疾人虚拟/增强现实系统。建议的研究计划包括本科和研究生水平的课程发展作为一个组成部分。它将有助于新的课程材料，涵盖跨学科的主题，从无线通信理论和数据驱动的强化学习到实时硬件系统原型。提出的IFAC方案的灵感来自于Viterbi在1965年的一个早期结果，该结果利用即时反馈来实现可靠的通信。在拟议的IFAC中，术语“瞬时”表示反馈延迟可以忽略不计，因为它明显短于平均符号长度。在Viterbi最初的工作中，瞬时反馈是一个理论假设，在当时的实际系统中是无法实现的。然而，拟议的研究旨在证明端到端实时系统的可行性，以表明由于数字集成电路技术的进步，自1965年以来实现了惊人的9个数量级（10亿倍）的改进，即时反馈确实是可能的。IFAC是一种无速率方案，它适应子比特级的信道实现，以优化频谱效率、错误率和/或给定目标中断概率的延迟。本研究计划将研究各种IFAC方案，这些方案是用特定应用的优化标准构建的，作为马尔可夫决策过程或强化学习问题，用于接收器的瞬时反馈决策。IFAC可以为大规模机器通信提供高度自适应的、基于在线学习的新方法，实现传统无线通信方案无法实现的卓越效率和性能。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。

This research program investigates instantaneous feedback-based adaptive communications (IFAC) for next generation wireless communication systems. The proposed IFAC enables sub-bit-level real-time adaptation of the wireless communication signal to exploit dynamic tradeoffs in spectral efficiency, reliability, and latency based on the instantaneous feedback from the receiver. Most conventional systems are designed for message-level link adaptation based on the average noise power without utilizing sub-bit-level noise observation. Such message-level feedback schemes are suboptimal for emerging and future applications that require ultra-reliable low latency communication (URLLC) in massive machine-type communication networks where the number of devices can be much greater than that of today's consumer mobile cellular networks. This research program will investigate the feasibility of instantaneous feedback-based wireless communication systems to attain a superior performance closer to the theoretical channel capacity and reliability limit, outperforming state-of-the-art conventional approaches. The proposed IFAC has great potential as a critical enabling technology for the next generation (beyond 5G) URLLC that targets 100 micro-second latency and 99.99999% reliability for short message transmission. Attaining such stringent URLLC goals with IFAC can enable various classes of new applications including autonomous micro-robots, intelligent automated factories, and virtual/augmented reality systems for people with disabilities. The proposed research program includes undergraduate- and graduate-level curricula development as an integral part. It will contribute to new course materials that cover interdisciplinary topics that range from the wireless communication theory and data-driven reinforcement learning to real-time hardware system prototyping.The proposed IFAC scheme is inspired by an early result by Viterbi in 1965 that utilizes instantaneous feedback for reliable communications. The term "instantaneous" in the proposed IFAC indicates that the feedback latency is negligible as it is significantly shorter than the average symbol length. In the original Viterbi's work, instantaneous feedback was a theoretical assumption that was not realizable in practical systems at that time. However, the proposed research aims at demonstrating the feasibility of end-to-end real-time systems to show that instantaneous feedback is indeed possible thanks to the advancement of digital integrated circuit technologies that attained astonishing 9 orders of magnitude (a factor of a billion) improvement since 1965. IFAC is a rate-less scheme that adapts to the channel realization at the sub-bit level to optimize the spectral efficiency, error rate, and/or the latency given the target outage probability. This research program will study various IFAC schemes that are constructed with an application-specific optimization criterion as a Markov decision process or reinforcement learning problem for the instantaneous feedback decision at the receiver. IFAC can enable highly adaptive, online learning based novel methods for massive machine-type communication, accomplishing superior efficiency and performance that are unattainable from conventional wireless communication schemes.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.