Unmanned aerial vehicles (UAVs) have become key tools in emergency missions in urban environments, due to their rapid deployment and ability to operate in hazardous conditions. Path planning is crucial in these disasters, as it determines the safety and efficiency of the mission. This study compares graph-based algorithms, such as A-Star (A) and D-Star Lite (D Lite), as well as sampling-based methods, including Probabilistic Roadmaps (PRM), Rapidly-exploring Random Tree Star (RRT), and Batch Informed Trees (BIT), along with spline-based post-processing applied to the trajectories produced by the sampling-based planners. To evaluate the performance of path planning algorithms, a framework developed in Python generates known three-dimensional urban environments with static obstacles. The benchmark considered 50 scenes, each with 15 obstacles, 5 UAVs, and 5 seed trials, resulting in 10,000 runs across eight different path planners. This study used a strict success criterion and analyzed planning time metrics, path length, minimum safety margin, and geometric smoothness. D¿ Lite achieved the highest success rate (98.8%) and the highest obstacle clearance, but at the cost of longer planning times. RRT¿ and RRT¿ with splines were the fastest methods (median approx 4.9 ms), although with lower success rates. In contrast, BIT¿ and BIT¿ with splines offered the best value, second only to D¿ Lite, combining high success rates with shorter paths, better path quality, and moderate computational demand. The benchmark provides a quantitative basis for planner selection in non-cooperative multi-agent deployment scenarios, without accounting for flight dynamics, sensor uncertainty, or dynamic obstacles.
