A multi-objective benchmark for UAV path planning with baseline results

초록

Unmanned Aerial Vehicles (UAVs) are increasingly deployed in complex environments for applications such as urban logistics, surveillance, and environmental monitoring. Path planning for UAVs in these settings is a multi-objective optimization problem that must balance conflicting criteria such as safety, efficiency, and regulatory compliance. While heuristic methods like A* and Rapidly-Exploring Random Tree (RRT) have been applied to this problem, they often struggle with convergence speed and solution quality in obstacle-rich, threedimensional spaces. Multi-Objective Evolutionary Algorithms (MOEAs) have shown strong potential for tackling such challenges, yet the absence of a comprehensive, standardized benchmark for UAV path planning continues to impede meaningful evaluation and comparison across studies. To address this gap, this paper makes three key contributions: (i) a parameterized framework for generating UAV path-planning problems across three representative environment types - urban, suburban, and mountainous - with adjustable scenario difficulty and real-world constraints such as no-fly zones; (ii) a curated benchmark suite comprising 14 diverse and rigorously designed test problems for reproducible and consistent algorithm evaluation; and (iii) baseline performance results for several state-of-the-art MOEAs, offering researchers clear reference points for future comparisons. By providing both a versatile testbed and a standardized evaluation methodology, this work aims to facilitate the development and fair assessment of UAV path-planning algorithms in realistic and challenging environments.

키워드