Innovative radar detection & AI classification techniques for small UAVs or drones

Patel, Jarez Satish (2024) Innovative radar detection & AI classification techniques for small UAVs or drones. PhD thesis, University of Glasgow.

Abstract

There is a growing interest in Micro-Drones among hobbyists, photographers, and corporations like Amazon, which plans to test drone delivery systems in the UK. However, these platforms have also been used for criminal activities, posing dangers such as trafficking and airspace disruptions. Radar sensors have shown promise in actively identifying and tracking these platforms, even in adverse conditions. Significant research is needed to understand their unique variability of signatures in order to facilitate successful detection and classification algorithmic implementations.
Research has shown that radar sensors can actively detect and assess such platforms. Furthermore, these sensors offer the advantage of operating effectively at long distances, in challenging weather conditions, and in low-light environments. To facilitate the safe integration of these new aerial vehicles into low-altitude airspace, further investigation is required to understand the distinctive signatures they emit, with radar systems being a good candidate to fulfill this.

This work discusses the development of a bespoke dual-band adaptive Frequency Modulated Continuous Wave (FMCW) radar system, distinguished by its unique ability to adjust numerous radar parameters in real-time, a capability seldom explored in academia. This project has been taken to a high Technology Readiness Level (TRL) and the system is capable of being deployed in the field without human intervention, whilst providing real-time signal processing and immediate user feedback. A research of the literature was firstly undertaken in order to identify challenges, the radar was concepted and progressed to manufacturing, it was tested to ensure that it fulfilled the criteria set out and then packaged into a portable enclosure. A novel signal processing implementation had to be developed from scratch in order to leverage the low power devices, whilst satisfying the problem definition. Through this functionality, the gathering of a large database of radar micro-Doppler signatures has been collected extensively over the course of two campaigns undertaken over Autumn 2023 and a classifier was designed, achieving a classification accuracy of 99.2% over 18K images.

This thesis is written in a manner as to guide a prospective radar engineer through the challenges presented in FMCW design and also the production lifecycle and deployement of a real system, together with the data processing constraints and the addressing of the original research goals.

• Item Type: Thesis (PhD)
• Qualification Level: Doctoral
• Subjects: T Technology > T Technology (General); T Technology > TK Electrical engineering. Electronics Nuclear engineering
• Colleges/Schools: College of Science and Engineering > School of Engineering > Electronics and Nanoscale Engineering
• Supervisor's Name: Bailey, Dr. Nicholas and Anderson, Dr. David
• Date of Award: 2024
• Depositing User: Theses Team
• Unique ID: glathesis:2024-84685
• Copyright: Copyright of this thesis is held by the author.
• Date Deposited: 12 Nov 2024 11:48
• Last Modified: 12 Nov 2024 11:56
• Thesis DOI: 10.5525/gla.thesis.84685
• URI: https://theses.gla.ac.uk/id/eprint/84685

Related URLs

• Conference proceeding
• Article DOI
• Article DOI
• Book section
• Article DOI
• Conference proceeding
• Conference proceeding
• Conference proceeding