Persistent aerial monitoring by autonomous, self-organized unmanned aerial vehicle swarms requires uninterrupted coverage, conflict-minimizing coordination, and map-consistent visual products quarriable in real time. We present a perception–mobility framework in which each unmanned areial vehicle follows a Eulerian patrol on a coverage graph, captures images geo-referenced, and fuses them on the fly into a global panorama (mosaic). After each patrol cycle, a vehicle selects a new seed vertex via dynamic graph restructuring that maximizes an information-gain objective over coverage deficit and mosaicking uncertainty, sustaining continuous monitoring and focusing sensing on poorly stitched regions. We jointly address: (I) continuous aerial monitoring under bandwidth/mobility constraints; (II) ground-based, real-time geo-referenced panorama construction; and (III) vertex-level conflicts from simultaneous arrivals. Our contributions are: Geo-referenced panorama pipeline with prior-guided alignment (global navigation satellite system/heading seeds), robust homography/pose estimation, and lightweight pose-graph optimization, with incremental seam optimization and exposure compensation. Adaptive, dynamic graph-restructuring algorithm that preserves Eulerian feasibility while reordering local traversals to minimize expected conflicts and bias motion toward high information gain. Mobility-aware transmission model that prioritizes keyframes, descriptors, and previews by their marginal impact on mosaic completeness. Simulations indicate higher mosaic completeness and reliability with reduced conflict incidence compared with perception-agnostic, fixed-order patrols.
Atashyan et al. (Sun,) studied this question.