Comparative Analysis of UAV Path Planning Algorithms Based on RRT for 3D Environment

Quadrotor Unmanned Aerial Vehicle (UAV) path planning, including global planning and local planning, has been widely studied by researchers in recent years. The rapidly exploring random tree (RRT) algorithm is one of the most popular methods for UAV path planning problems. Various path-planning algorithms have been developed based on RRT in the last two decades, such as RRT*, RRT*-Smart, Informed-RRT*, and connect methods that grow two trees at the same time, which are normally considered to be faster and more efficient than regular RRT, including RRT-Connect, RRT*-Connect, and Informed-RRT*-Connect. All the RRT variants can be used for global path planning in 3D space for UAVs. The performance of each algorithm varies in different environments, and the result of planning in 3D space also differs from the result in 2D space. This paper briefly reviews each RRT variant's features and presents algorithm performance evaluated in different featured environments. Indicators such as convergence time, initial path length, final path length, and path length per iteration are recorded. Multiple tests are conducted, and the average value of the indicators is compared due to the non-consistency performance feature of RRT-based algorithms. The connect methods are not always faster than the regular RRT variants. Each algorithm has its best-performance scenario. Furthermore, the characteristics and limitations of each algorithm in featured environments are concluded and can help with path-planning algorithm selection in specific application scenarios.