NSF-IITP: AI/ML-Enabled Scalable and Privacy-Preserving 6G Space-Air-Ground Integrated Network Operation

第六代(6G)无线技术预计将提供跨越人类、机器和传感器的超级连接，推动扩展现实(XR)、人工智能(AI)和自主机器人等领域令人兴奋的新应用的增长。该项目的重点是关键的使能技术，即综合非地面网络(NTN)，它除了传统的地面站外，还包括空间、空中和地面组件，如低地球轨道(LEO)卫星、高空平台站(HAPS)和无人驾驶飞行器(UAV)。从固定位置解放出来的天地一体化网络可以支持与偏远地区(例如气候监测)、灾区、热点和覆盖洞以及高机动性集群(如飞机和船只)的无缝连接。然而，如此灵活的网络架构带来了重大的技术挑战。该项目致力于通过利用强大的人工智能和机器学习(ML)技术来探索NTN关键操作问题的新解决方案。值得注意的是，拟议的研究旨在受益于参与研究的美国和韩国机构之间的密切合作。研究成果将大大推进6G集成网络的理论和实践，确保美国/韩国劳动力的全球技术领先地位，并通过提供关键基础设施来解决其中的关键问题，为社会和环境议程做出贡献。所获得的知识也将对其他科技领域产生影响，如网络科学、数据科学、分布式机器人和保护隐私的智能健康。随之而来的教育部分将提供适合在美国和韩国培养STEM人才的新鲜学习体验。更具体地说，该项目旨在解决与NTN运营相关的关键挑战，从无线电环境分析、空间-空中-地面综合布线、多卫星协调、服务感知资源分配，到隐私保护。虽然AI/ML方面的最新进展预计将是这一努力的适当推动因素，但据观察，为了确保AI/ML模块的培训和操作的效率和健壮性，传统的数据驱动的黑盒方法需要得到证实的特定领域范例和新颖的ML体系结构见解的补充。在这种背景下，ML、信号处理、通信、网络和信息理论方面的不同专业知识将通过紧密的国际合作汇聚在一起，做出变革性的贡献。要探索的重要研究议程包括：1)通过联合ML和信号处理来绘制复杂的无线电环境地图；2)通过约束多智能体强化学习来实现可扩展的网络优化；以及3)在无线网络上进行隐私保护推理的基本权衡。此外，将对这些议程进行综合研究，为高度动态的NTN场景和保护隐私的MAP推理和多代理协调方案设计MAP辅助网络控制方法。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。

The 6th generation (6G) wireless technology is envisaged to provide hyperconnectivity across humans, machines, and sensors, fueling the growth of exciting new applications in expanded reality (XR), artificial intelligence (AI), and autonomous robotics, to name a few. This project focuses on the key enabling technology, namely, the integrated non-terrestrial networks (NTNs), which encompass space, air, and ground components, such as the low Earth orbit (LEO) satellites, high-altitude platform stations (HAPSs), and unmanned aerial vehicles (UAVs), in addition to the traditional terrestrial stations. Freeing itself from fixed locations, the space-air-ground integrated network can support seamless connectivity to remote regions (e.g. for climate monitoring), disaster areas, hot spots, and coverage holes, as well as high-mobility clusters such as aircrafts and vessels. Significant technical challenges emerge, however, with such a flexible network architecture. This project aspires to explore novel solutions to critical operational issues of NTNs, by tapping into powerful AI and machine learning (ML) techniques. Notably, the proposed research is designed to benefit from close collaboration among the participating US and South Korean institutions. The research outcomes will substantially advance the theory and practice of 6G integrated networking, secure global technological leadership of the US/Korean workforce, and contribute to societal and environmental agenda by providing vital infrastructure to combat the critical issues therein. The gained knowledge will have impact to other science and technology domains as well, such as network science, data science, distributed robotics, and privacy-preserving smart health. The attendant educational components will provide fresh learning experiences suitable for preparing STEM talents in the US and South Korea.More specifically, the project aims at addressing key challenges associated with NTN operation, ranging from radio environment analysis, space-air-ground integrated routing, multi-satellite coordination, service-aware resource allocation, to privacy protection. While recent advances in AI/ML is expected to be the opportune enabler for this endeavor, it is observed that to ensure efficiency and robustness in the training and operation of the AI/ML modules, traditional data-driven black box approaches need to be complemented with proven domain-specific paradigms and novel ML architectural insights. In this context, diverse expertise in ML, signal processing, communication, networking, and information theory will be pooled together through tight international collaboration to make transformative contributions. Important research agenda to be explored include: 1) Complex radio environment cartography through joint ML and signal processing; 2) scalable network optimization via constrained multi-agent reinforcement learning; and 3) fundamental trade-offs in privacy-preserving inference over wireless networks. Furthermore, integrative research of these agenda will be pursued to devise map-assisted network control methods for highly dynamic NTN scenarios and privacy-preserving map inference and multi-agent coordination 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.