Kinodynamic Motion Planning for Holonomic UAVs in Complex 3D Environments

Abstract

Due to their ability to hover, rotorcraft can be used in many cluttered and narrow environments that are unsuitable for other unmanned aerial vehicles (UAVs). However in such environments, it is usually difficult to find paths that are both collision-free and dynamically feasible. This paper introduces a computationally efficient algorithm for finding safe paths, through known static three dimensional environments, that satisfy the kinodynamic constraints of a quadrotor. First, a collision-free path that ignores kinodynamic constraints is found using probabilistic roadmaps (PRM). Next, a heuristic re-sampling algorithm is used to make improvements to this path. Finally, the path is separated into a series of motion primitives that satisfy kinodynamic constraints. To ensure corner path segments remain collision-free, a method of bounding the vehicle deviation from the piece-wise linear path at each corner is introduced. Finally, the control inputs required to traverse the transformed path are calculated using kinematics and become the feed-forward inputs for a position controller. In simulation, the algorithm is able to successfully compute safe kinodynamic paths through cluttered and narrow environments, while using limited computational resources.