Cooperative Multi-UAV Task Allocation and Smoke Deployment for Simulation-Based Line-of-Sight Screening

Abstract

This study examines how multiple uncrewed aerial vehicles (UAVs) should allocate tasks and coordinate smoke deployment to interrupt line of sight under fast, spatially dispersed incoming threats. A three-dimensional simulation framework is developed that couples proportional-navigation-like reference threat motion, aerodynamic drag, overload saturation, field-of-view break-lock and reacquisition, smoke-puff advection–diffusion–settling, and infrared/electro-optical attenuation through the Beer–Lambert law. Under the same five-UAV physical model and search budget, we compare a non-optimized greedy estimated-time-of-arrival heuristic (M0) with three genetic-algorithm-coupled task-organization structures: round-robin balancing (M1), spatiotemporal clustering and matching (M2), and capacity-constrained global preallocation (M3). Four topologies are evaluated with 30 independent runs per topology and 20 threat tracks per run. The aggregate successful-screening rates are 0.00%, 34.96%, 65.83%, and 81.25% for M0–M3, respectively; in the most dispersed omnidirectional topology, the corresponding rates are 0.00%, 41.50%, 29.83%, and 66.50%. Run-level tests confirm the overall M2–M3 separation (Welch p = 0.0004603; Mann–Whitney p = 0.006083). Component checks, threshold scans, optimal-transport-invariance tests, one-factor-at-a-time sensitivity analysis, and runtime benchmarks indicate that M2 is limited by structural coverage loss in dispersed settings, whereas M3 preserves task admission more stably. The study is positioned as a simulation-based mechanism comparison rather than field validation.