Comparison between A*and RRT Algorithms for 3D UAV Path Planning

Abstract

This paper aims to present a comparative analysis of the two most utilized graph-based and sampling-based algorithms and their variants, in view of 3D UAV path planning in complex indoor environment. The findings of this analysis outline the usability of the methods and can assist future UAV path planning designers to select the best algorithm with the best parameter configuration in relation to the specific application. An extensive literature review of graph-based and sampling-based methods and their variants is first presented. The most utilized algorithms which are the A*for graph-based methods and Rapidly-Exploring Random Tree (RRT) for the sampling-based methods, are defined. A set of variants is also developed to mitigate with inherent shortcomings in the standard algorithms. All algorithms are then tested in the same scenarios and analyzed using the same performance measures. The A*algorithm generates shorter paths with respect to the RRT algorithm. The A*algorithm only explores volumes required for path generation while the RRT algorithms explore the space evenly. The A*algorithm exhibits an oscillatory behavior at different resolutions for the same scenario that is attenuated with the novel A*ripple reduction algorithm. The Multiple RRT generated longer unsmoothed paths in shorter planning times but required more smoothing over RRT. This work is the first attempt to compare graph-based and sampling-based algorithms in 3D path planning of UAVs. Furthermore, this work addresses shortcomings in both A*and RRT standard algorithms by developing a novel A*ripple reduction algorithm, a novel RRT variant and a specifically designed smoothing algorithm.