Operator-Theoretic and Data-Driven Approaches to Radar Modeling and Signal Processing
Open Access
- Author:
- Pici, Caden
- Graduate Program:
- Electrical Engineering
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- March 30, 2023
- Committee Members:
- Constantino Lagoa, Major Field Member
Ram Narayanan, Chair & Dissertation Advisor
Tim Kane, Major Field Member
Sastry Kompella, Special Member
Asok Ray, Outside Unit & Field Member
Madhavan Swaminathan, Program Head/Chair - Keywords:
- radar
signal processing
Koopman
sea clutter - Abstract:
- Optimizing radar signal processing algorithms is typically addressed under various assumptions about the nature of the targets, clutter, and noise. This can include target motion and signature, clutter distribution, and Gaussian nature of the noise. In this work, a generalized linear operator based model of the radar problem is presented which combines aspects of a couple different works in literature, and overcomes the limitations present in the linear time-invariant model often used. From the general equation for monostatic radar, connections are made to the work presented in the remainder of the thesis, in which the following chapters study different aspects of that equation under different assumptions and conditions. This includes a filtering method for adapting noise radar waveforms to the eigenfunction solution of maximizing the signal-to-interference-plus-noise ratio in the presence of colored noise. The optimal solution to this problem is limiting in that it requires knowledge of the target impulse response. Two target classification receivers are presented which rely on having a library of target response data across different aspect angles. One solution is based on performing a sparse regression onto the space of possible frequency responses. Though providing the best accuracy, this is a slower algorithm. This is addressed by providing a target-tailored matched filter band which is shown to have promise in classifying targets in the simulation study. Finally, an existing stochastic differential equation model of sea clutter is reviewed, and a data-driven framework relying on Koopman operator theory and dynamic mode decomposition is used to discover a model of sea clutter dynamics. This model generates a higher dimensional linear operator that is able to reasonably well model the temporal dynamics of the sea clutter stochastic differential equations. From this model, an anomaly detection approach is studied, which is able to detect changes in the sea state, and shows potential for future work on small target detection in the presence of sea clutter.