Synthetic aperture imaging by distributed arrays of space telescopes

We present synthetic aperture imaging by distributed arrays of space telescopes (SAIDAST), a methodology for high-resolution space imaging using distributed arrays of small telescopes. SAIDAST reconstructs images from sparse, motion-blurred intensity measurements without requiring large monolithic apertures. We compare two implementations using numerical simulations: Mo-PIE (motion-aware ptychography) and a physics-informed neural network. The neural network demonstrates high reconstruction quality (SSIM up to 0.985 on a USAF 1951 resolution target and up to 0.85 on Europa JunoCam imagery), graceful degradation at negative SNRs under an intensity-dependent Gaussian shot-noise proxy with parity to the Mo-PIE baseline above 0 dB, and constant-memory sequential processing across 64-256 detectors on a single consumer-grade GPU. Simulations with 0.17 m apertures forming a 10 m synthetic array validate the approach: on the USAF 1951 target, SAIDAST reaches 87.7% of ground-truth normalized MTF-AUC versus under 1% for a single 17 cm aperture, enabled by the combination of expanded k-space coverage and physics-informed DIP reconstruction. A paired partial-coherence simulation (n = 4, Welch p = 0.025) indicates an approximately 24% SSIM degradation on Europa when the coherent forward-model approximation is relaxed. Based on SAIDAST, we propose that a distributed array of satellites can form a high-resolution telescope in the visible spectral region.