Cooperative multi-UAV smoke-screen deployment for visibility denial: a spatio-temporal optimization approach

Abstract Spatio-temporal smoke-release planning for line-of-sight shielding is nontrivial because feasible schedules must satisfy coupled UAV kinematics, fuse delays, and visibility-dependent screening effects under expensive simulation-based evaluation. This study formulates a unified scene model that couples a risk–pressure exposure layer (RPEL) with sightline-coverage screening (SCS), and proposes SPARROW (Spatio-temporal Pressure–Aware Release Optimization). SPARROW builds per-UAV candidate libraries on a discretized spatio-temporal grid, composes multi-device schedules under feasibility constraints, and applies a shared fine-grid refinement for reporting. The framework is evaluated on four operational tasks (P1–P4) and a 27-scenario scalability grid parameterized by the numbers of UAVs, smoke devices per UAV, and protected platforms. Comparative benchmarking against genetic algorithm (GA), particle swarm optimization (PSO), and continuous ant colony optimization (ACO) uses an identical evaluator and an equal budget of 1092 coarse plus 20 fine objective evaluations per run, with 8 independent runs per scenario. Across the 27 scenarios, SPARROW attains an average effective shielding duration ( J step ) of 4.44 ± 0.33 s (success 81.9%, feasibility 98.6%), compared with PSO (3.49 ± 1.30 s), GA (1.99 ± 1.34 s), and ACO (0.41 ± 0.39 s); the integrated shielding metric is J area = 4.501 ± 0.340 for SPARROW. A proxy sensitivity study in P4 with wind/turbulence perturbations and a soft-transmittance variant preserves the main performance trends. The results provide an optimization-ready and reproducible pipeline for multi-UAV smoke-release scheduling and controlled evidence for budgeted benchmarking of stochastic optimizers for visibility screening.