Distributed small-satellite constellations in Low Earth Orbit (LEO) enable global radio-frequency sensing, but accurate localization of non-cooperative emitters without waveform knowledge remains challenging. Existing time- and frequency-difference methods can only achieve kilometer-scale accuracy for narrowband signals. These methods require inter-satellite baselines of tens of kilometers and suffer from spatial ambiguities when extended to phase-coherent processing. This dissertation presents a unified framework for carrier-phase-based emitter geolocation using distributed CubeSat formations. This framework improves performance for both narrowband and wideband signals. The central contribution is a geometry-accurate direct position determination architecture that exploits carrier-phase coherence across spatially separated satellites. Unlike classical two-stage approaches, emitter position is estimated directly from the received signals, enabling sub-wavelength path-length sensitivity. A Minimum Variance Distortionless Response (MVDR) beamforming formulation provides robust localization, extended to wideband signals via short-time Fourier transform processing that leverages frequency diversity to suppress ambiguities. For narrowband maritime signals at 162 MHz, sub-kilometer two-dimensional accuracy is achieved, while wideband signals at 2.8 GHz enable sub-hundred-meter accuracy under favorable conditions. Bayesian Cramér–Rao bounds are derived under realistic phase noise and satellite ephemeris uncertainty, revealing fundamental accuracy floors dominated by synchronization and orbit determination errors at high signal-to-noise ratios. Large inter-satellite baselines, while improving angular resolution, introduce spatial aliasing; this work analyzes mitigation via local phased arrays, fusion with time- and frequency-difference measurements, and exploitation of frequency and temporal diversity. For multi-emitter scenarios, a hybrid neural-network pipeline processes beamforming-based position likelihood maps to jointly estimate emitter count and locations. This approach achieves emitter-count accuracies exceeding 95% above −5 dB SNR and reduces localization error by up to 40% compared to classical peak detection. Overall, the results demonstrate that phase-coherent CubeSat formations can achieve order-of-magnitude improvements in localization accuracy over conventional methods using baselines of only tens to hundreds of meters, enabling autonomous space-based spectrum monitoring with precision previously requiring large platforms or dense ground infrastructure.     «

Distributed small-satellite constellations in Low Earth Orbit (LEO) enable global radio-frequency sensing, but accurate localization of non-cooperative emitters without waveform knowledge remains challenging. Existing time- and frequency-difference methods can only achieve kilometer-scale accuracy for narrowband signals. These methods require inter-satellite baselines of tens of kilometers and suffer from spatial ambiguities when extended to phase-coherent processing. This dissertation presents...     »