CNS Core: Small: Robust and Decentralized Mobility and Connectivity Management in UAV-Enabled Aerial Wireless Networks

随着技术的进步，无人机变得更轻、更便宜、更容易部署。结合其机动性、自主性和灵活性，无人机现在已准备好用于各种现实世界场景(如运送医疗用品、救灾、紧急搜救、环境监测和遥感)，并有望成为下一代无线网络的组成部分。与此同时，将自动无人机集成到下一代无线网络中，同时保持可靠性、安全性和连接性是一项关键任务，仍然存在许多挑战和问题。在这方面的几个关键问题如下。空域限制(例如，安全和避免碰撞)将如何与网络连接要求相互作用？随着无人机数量的增长和集中控制变得令人望而却步，能否在空中网络中实现移动性和连通性管理的去中心化自主决策？能否制定分散/分散的资源分配政策？这些政策将如何影响移动性和连通性管理？决策政策对对抗性攻击的敏感性是什么？在存在敌意攻击的情况下，能否在空中网络中实现稳健的学习和安全的决策？因此，存在一个关键的知识鸿沟，该项目通过相互关联的推进来解决这一鸿沟。具体地说，本项目的主要目标是设计和分析支持无人机的无线网络中稳健、智能和分散的决策策略。为了实现自主空中网络的愿景，在具有挑战性但实际上更相关的分散操作环境中，移动性和连通性管理的关键考虑因素与动态资源分配被结合在一起。在这种情况下，无人机路径规划中同时考虑了连通性和控制约束，如碰撞避免要求和运动学限制。分散和自主的决策战略将通过仅在当地观察/感知的分布式学习来制定。决策的稳健性将通过对对抗性攻击的敏感性分析和防御机制的发展来建立。总体而言，该项目寻求通过结合无线网络、优化、控制、深度强化学习和对抗性学习的工具来建立强大的分析和算法框架。这项研究的成果将为救灾和遥感等应用提供支持。这项研究项目与教育的整合将通过新引入的课程来完成，并让本科生参与PIS的研究计划。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。

With the advances in the technology, unmanned aerial vehicles (UAVs) have become lighter, cheaper, and easier to deploy. Combined with their mobility, autonomy and flexibility, UAVs are now poised to be used in a variety of real-world scenarios (such as delivery of medical supplies, disaster relief, emergency search and rescue, environment monitoring, and remote sensing), and are expected to be an integral part of next-generation wireless networks. At the same time, integration of autonomous UAVs into next-generation wireless networks while maintaining the reliability, safety, and connectivity is a critical undertaking, and many challenges and questions remain. Several key questions in this context are the following. How would airspace constraints (e.g., safety and collision avoidance) interact with network connectivity requirements? As the number of UAVs grows and centralized control becomes prohibitive, can decentralized autonomous decision-making be achieved for mobility and connectivity management in aerial networks? Can decentralized/distributed resource allocation policies be developed and how would these policies impact mobility and connectivity management? What is the sensitivity of decision-making policies to adversarial attacks? Can robust learning and secure decision-making be achieved in aerial networks in the presence of adversarial attacks? Hence, there exists a critical knowledge gap, and this project addresses this gap via interconnected thrusts. Specifically, the primary objective in this project is to design and analyze robust, intelligent, and decentralized decision-making policies in UAV-enabled wireless networks. To realize the vision of autonomous aerial networks, the critical considerations of mobility and connectivity management are addressed jointly with dynamic resource allocation in the challenging but practically more relevant setting of decentralized operation. In this setting, both connectivity and control constraints such as collision avoidance requirements and kinematic limitations in UAV path planning are taken into account. Decentralized and autonomous decision-making strategies are to be developed via distributed learning with only local observation/sensing. Robustness in decision-making is to be established via sensitivity analysis to adversarial attacks and the development of defensive mechanisms. Overall, the project seeks to establish a strong analytical and algorithmic framework by combining tools from wireless networking, optimization, control, deep reinforcement learning, and adversarial learning. The outcomes of this research will support applications such as disaster relief and remote sensing. Integration of this research project with education will be accomplished through newly introduced courses, and involving undergraduates in the PIs' research program.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.