SWIFT-SAT: Network Adaptation Based on Physics-Inspired Learning Framework for Radio Coexistence of Terrestrial and Satellite Information Systems

信息技术的重大进步刺激了许多新概念和无线创新，包括3GPP的5G地面和非地面网络（NTN）服务。随着来自人工智能、自动化、物联网和虚拟/增强现实等大量应用的大量数据流量继续涌入现有容量有限的系统，不断增长的频谱需求促使无线网络扩展到新的无线电频段。随着地面无线电频谱的不断扩大，地面和卫星系统的频谱共存变得越来越重要。这一SWIFT-SAT合作项目旨在通过整合物理无线电传播模型和数据驱动学习机来开发物理启发的学习框架，以促进地面和卫星系统之间高效、和谐的共存。利用来自测量数据的先验知识和无线电物理模型的约束，拟议的研究活动旨在建立准确的无线电干扰模型和数据库，以实现先进的网络优化和适应。准确的无线电覆盖图估计是地面系统与卫星系统共存干扰的关键。主要项目重点是无线电地图的动态估计和预测，以及在共存约束下将其集成到无线网络优化和自适应中。研究人员只需要在空间、频率和时间域中进行稀疏观测，采用物理启发和模型驱动的学习方法，基于共存无线系统的无线电地图估计来准确估计同信道和相邻信道的干扰。该项目的创新进一步包括以传播模型为指导的基于学习的方法，用于智能无线网络优化和无线电地图信息促进的干扰诊断。通过提出的创新，基于模型的无线电覆盖估计可以预测干扰异常，可以快速检测或诊断网络中断，并可以有效地响应网络适应策略。拟议的解决方案和项目成果预计将对地面和卫星系统之间干扰有限共存的未来技术发展产生重大影响。更广泛地说，这项工作促进了跨数据采集、通信、传感和分布式计算的可靠和智能无线系统。这个项目更广泛的影响包括这项研究激发了许多新的教育机会。项目团队已经与当地的市中心学校合作，计划将K-12和代表性不足的人才在STEM方面的学习机会纳入该项目。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。

Major advances in information technologies have stimulated numerous novel concepts and wireless innovations including 3GPP’s 5G terrestrial and non-terrestrial network (NTN) services. As high volumes of data traffic from a rich plethora of applications ranging from artificial intelligence, automation, IoT, and virtual/augmented reality continue to cram into existing system of limited capacity, the ever-increasing spectrum need motivates the expansion of wireless networking into new radio bands. As the terrestrial radio spectrum expands, spectrum coexistence between terrestrial and satellite systems becomes increasingly critical. This collaborative SWIFT-SAT project aims to develop physics-inspired learning frameworks by integrating physical radio propagation models and data-driven learning machines to facilitate efficient and harmonious coexistence between terrestrial and satellite systems. Leveraging prior knowledge from measurement data and the constraints imposed by radio physical models, the proposed research activities are geared towards establishing accurate radio interference models and databases to empower advanced network optimization and adaptation. Accurate radio coverage map (radio-map) estimation is critical to interference-limited coexistence between terrestrial and satellite systems. Major project thrusts are directed at dynamic estimation and prediction of radio-map as well as their integration into wireless network optimization and adaptation under coexistence constraints. Requiring only sparse observations in spatial, frequency, and temporal domains, the investigators employ physics-inspired and model-driven learning approaches to accurately estimate co-channel and adjacent channel interferences based on radio-map estimation of coexisting wireless systems. The project innovation further includes learning-based approaches guided by propagation models for intelligent wireless network optimization and interference diagnosis facilitated by radio-map information. Through the proposed innovation, model-based radio coverage estimation can anticipate interference anomalies, can detect or diagnose network outages quickly, and can respond with effective network adaptation policies. The proposed solutions and project outcomes are expected to significantly impact future technology development for interference-limited coexistence between terrestrial and satellite systems. More broadly, this work promotes reliable and intelligent wireless systems across data acquisition, communication, sensing, and distributed computing. Broader impacts from this project include many new educational opportunities stimulated by this research. Already in partnership with local inner-city schools, the project team plans to incorporate into this project learning opportunities in STEM for K-12 and under-represented talents.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.