用于全球导航卫星系统电子干扰检测与分类的人工智能模型

ARTIFICIAL INTELLIGENCE MODELS FORELECTRONIC INTERFERENCE DETECTION ANDCLASSIFICATION ON GLOBAL NAVIGATIONSATELLITE SYSTEMS Nordgauer, Brynley J. Monterey, CA; Naval Postgraduate School https://hdl.handle.net/10945/74156 This publication is a work of the U.S. Government as defined in Title 17, UnitedStates Code, Section 101. Copyright protection is not available for this work in theUnited States. Downloaded from NPS Archive: Calhoun NAVALPOSTGRADUATESCHOOL MONTEREY, CALIFORNIA THESIS THIS PAGE INTENTIONALLY LEFT BLANK Global navigation satellite systems, such as GPS, transmit position, navigation, and timing data over theelectromagneticspectrum.These transmissions are susceptible to interference via malicious actorsperforming jamming or spoofing on them. Due to the remote nature of space, direct human oversight is notalways possible; therefore, automated detection and classification methods are needed. Automated detectionand classification of interference can give end-users better awareness about the integrity of the data theyreceive and allow them to take appropriate countermeasures if needed. This thesis investigated the use ofartificialintelligence,specifically deep learning architectures,to provide automated detection andclassification of electronic interference for use on resource-constrained flight hardware. The main challengewith these models is that they typically require substantial computer resources. These resources are notavailable on small satellite flight computers, so an efficient model architecture is needed. Ultimately, a denseautoencoder architecture was found to have the best balance between performance and storage requirement.Thisthesis demonstrated that lightweight models can automatically detect and classify electronicinterference and provides methods for integrating these models into flight hardware. The methods developedin this research are aimed at satellite developers who wish to provide reliable data to their end-users. THIS PAGE INTENTIONALLY LEFT BLANK Distribution Statement A. Approved for public release: Distribution is unlimited. ARTIFICIAL INTELLIGENCE MODELS FOR ELECTRONICINTERFERENCE DETECTION AND CLASSIFICATION ON GLOBALNAVIGATION SATELLITE SYSTEMS Brynley J. NordgauerEnsign, United States NavyBS, United States Naval Academy, 2024 Submitted in partial fulfillment of therequirements for the degree of MASTER OF SCIENCE IN ASTRONAUTICAL ENGINEERING from the NAVAL POSTGRADUATE SCHOOLJune 2025 Approved by:Mark KarpenkoAdvisor Wenschel D. LanCo-Advisor Garth V. HobsonChair, Department of Mechanical and Aerospace Engineering THIS PAGE INTENTIONALLY LEFT BLANK ABSTRACT Global navigation satellite systems, such as GPS, transmit position, navigation,and timing data over the electromagnetic spectrum. These transmissions are susceptibleto interference via malicious actors performing jamming or spoofing on them. Due to theremote nature of space, direct human oversight is not always possible; therefore,automated detection and classification methods are needed. Automated detection andclassification of interference can give end-users better awareness about the integrity ofthe data they receive and allow them to take appropriate countermeasures if needed. Thisthesisinvestigated the use of artificial intelligence,specifically deep learningarchitectures, to provide automated detection and classification of electronic interferencefor use on resource-constrained flight hardware. The main challenge with these models isthat they typically require substantial computer resources. These resources are notavailable on small satellite flight computers, so an efficient model architecture is needed.Ultimately, a dense autoencoder architecture was found to have the best balance betweenperformance and storage requirement. This thesis demonstrated that lightweight modelscan automatically detect and classify electronic interference and provides methods forintegrating these models into flight hardware. The methods developed in this research areaimed at satellite developers who wish to provide reliable data to their end-users. THIS PAGE INTENTIONALLY LEFT BLANK Table of Contents 1Introduction1 1.1Problem Statement........................11.2Research Questions.......................11.3Research Approach.......................21.4Research Context........................31.5Research Limitations .......................41.6Thesis Outline.........................4 2Theory 7 2.1Chapter Introduction .......................72.2Artificial Intelligence and Machine Learning.............72.3Programming Tools.......................112.4Data Description........................122.5Electronic Interference ......................172.6Data Formatting.........................192.7Machine Learning ........................212.8Resource Constraints .......................232.9Chapter Summary ........................23 3Methodology 25 3.1Chapter Introduction .......................253.2Data Handlin