Date of Award
Spring 2026
Language
English
Embargo Period
5-1-2028
Document Type
Master's Thesis
Degree Name
Master of Science (MS)
College/School/Department
Department of Electrical and Computer Engineering
Program
Electrical and Computer Engineering
First Advisor
Bariscan Yonel
Committee Members
Nathan Dahlin, Dola Saha
Keywords
Incoherence, compressed sensing, waveform diversity, adaptivity
Subject Categories
Controls and Control Theory | Electrical and Computer Engineering | Engineering | Signal Processing
Abstract
The thesis explores the utilisation of multistatic systems (multiple transceiver nodes) for wave-based imaging, with a focus on the design of algorithms capable of reconstructing scenes from limited measurements. The proposed approach is evaluated across varying numbers of transceivers and signal-to-noise ratio (SNR) levels to assess its performance. The work addresses limitations of existing methods, including MIMO systems based on sequential protocols with fixed waveforms, as well as approaches such as radar coincidence imaging, which offer rich sensing capabilities at the expense of increased computational and implementation complexity.
We design alternative algorithms with one clear aim: to exploit the improved spatial diversity of distributed systems while reducing reliance on fixed illumination strategies. In contrast to conventional methods that employ static sensing patterns, we formulate adaptive imaging as an online bilinear problem and introduce a novel incoherence prior, enabling the design of an incoherent measurement ensemble in an online manner.
Through an alternating optimisation scheme, we update the transmit state s and reflectivity ρ sequentially. At each pulse, the imaging problem admits a rank-one structure, allowing efficient updates while maintaining flexibility across different sensing configurations.
The framework accommodates both static and waveform-diverse settings and naturally extends to scenarios involving moving agents.
For the forward model, we assume single scattering under the Born approximation, with sequential updates performed per pulse. We further introduce signed variants of the prior to control the trade-off between exploration and exploitation in the two-dimensional Fourier domain. In particular, waveform diversity introduces natural exploration of the measurement subspace, allowing the algorithm to emphasise exploitation for improved local resolution,whereas in static settings exploration must be actively induced through the transmit state.
Experimental results demonstrate that the proposed algorithm achieves improved reconstruction performance compared to baseline strategies, including fixed single-transmitter operation, cyclic activation, and data-driven state selection. The method exhibits faster convergence and maintains robustness across a range of operating regimes, defined by varying transceiver counts and SNR levels.
The work establishes a fundamental connection between waveform diversity and adaptive imaging, showing that measurement diversity can be introduced either actively through state design or passively through waveform variation, with distinct implications for reconstruction performance and convergence behaviour.
License
This work is licensed under a Creative Commons Attribution 4.0 International License.
Recommended Citation
Membe, Keith, "Online Learning for Adaptive Distributed Wave Based Imaging" (2026). Electronic Theses & Dissertations (2024 - present). 452.
https://scholarsarchive.library.albany.edu/etd/452