CNS Core: Small: PilotPC: Proactive Inverse Learning of Network Topology for Predictive Communication among Unmanned Vehicles

在当今的许多应用中，无人机的使用已经变得司空见惯。与许多其他技术一样，无人机在带来威胁和安全问题的同时，也为许多应用中的当前实践带来了舒适性和效率。该项目将在提供隐私和安全的同时，在开发更安全、更可靠的空中监测平台方面推进科学前沿。为了实现这一目标，该项目旨在通过使用反向学习技术，在无人机网络中做出准确的预测，从而实现更安全和更高效的操作。将开发一种智能观察系统，以持续监控自主代理的行为，通过测量其在给定条件下偏离最合理的行为来识别可疑活动。研究包括以下几个步骤：(I)建立机制来模拟目标节点对周围环境的感知(即从目标的眼睛看到单词)，(Ii)通过反向学习(读取目标的大脑)来模拟目标节点的决策过程，预测目标节点在给定条件和可用资源下的理性行为(S)，以及(Iii)通过检测罕见但关键的偏离预期理性行为的事件来衡量信任水平。特别是，无人机的移动性及其联网活动将作为恶意活动的两个重要场景进行调查。该项目预计将导致开发新的工具和算法，用于智能节点的行动分析，并将与研究社区共享。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 use of unmanned aerial vehicles (UAVs) has become commonplace in many applications nowadays. Like many other technologies, UAVs brings comfort and efficiency to the current practice in many applications while posing threats and safety issues. This project will advance the frontier of science in developing safer and more reliable aerial monitoring platforms while offering privacy and safety. Towards this goal, this project aims to make accurate predictions in a network of unmanned aerial vehicles, by using inverse learning techniques, so as to enable safer and more efficient operations.A smart observation system will be developed to constantly monitor the behavior of autonomous agents to identify suspicious activities by measuring their deviation from the most rational behavior under given conditions. The proposed research includes several steps: (i) to develop mechanisms to simulate the target node's perception of its surrounding environment (i.e., seeing the word from the target's eyes) using deep learning methods, (ii) to predict its rational behavior(s) under given conditions and available resources by modeling the target's decision making procedure through inverse learning (reading the target's brain), and (iii) to measure the level of trust by detecting rare but critical episodes of deviating from its expected rational behavior. In particular, the mobility of UAVs and their networking activities will be investigated as two important scenarios of malicious activities. This project is expected to result in developing new tools and algorithms for the action analysis of smart nodes that will be shared with the research community.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.