Título: Simulation and Analysis of UAV Safe Landing Methods in Urban Scenarios
Autores: Marcelo Vieira Santos, Erickson Nascimento
Resumo: The increasing deployment of unmanned aerial vehicles (UAVs) in urban environments necessitates robust systems to enable safe emergency landings amidst crowded conditions. This study addresses key gaps in existing approaches by developing a comprehensive simulation framework using Unreal Engine and AirSim. The framework allows systematic variation of critical parameters such as crowd density, people speed, lighting, image corruption and data processing latency, providing a robust foundation for evaluating UAV safe landing systems. Safety map generation methods based on crowd density-based maps and people localization maps were analyzed to explore their trade-offs. The density-based approach demonstrated superior robustness to motion blur, maintaining usable performance under moderate corruption levels. However, it faced challenges with higher abort rates and longer mission durations. The localization-based method had better performance in normal visual conditions, achieving higher success rates across varying crowd densities, speeds and data processing latencies. This method also showed faster mission completion times in all experiments. We conclude that the density-based method is more suitable for harsh visual conditions, while the localization-based method is preferable for normal conditions or when timing is critical. Future work should focus on hybrid methods that dynamically adapt to the environment, real-world validation and alignment with regulatory standards to ensure deployment readiness. These advancements will enable safer UAV operations in urban settings.
Palavras-chave: Autonomous UAV Safe Landing; Robust UAV Systems; Urban Environments; Crowd Density; Crowd Localization; Distance Map; Simulation; AirSim.
Páginas: 8
Código DOI: 10.21528/CBIC2025-1174195
Artigo em PDF: CBIC_2025_paper1174195.pdf
Arquivo BibTeX:
CBIC_2025_1174195.bib