RRT*アルゴリズムによるUAV写真測量ミッションでの衝突回避
Utilizing the RRT*-Algorithm for Collision Avoidance in UAV Photogrammetry Missions
写真測量ミッション中のマルチコプタが障害物を回避しつつ元の経路に復帰できるようRRT*を適用し、Gazebo/PX4シミュレーションで最大6m/sまで検証した。
著者: Lars Killian, Jan Backhaus
分類: cs.RO
原文アブストラクト
This paper presents the application of the Rapidly-exploring Random Tree Star (RRT*) algorithm for multicopter collision avoidance in photogrammetry missions. For better applicability, the presented algorithm redirects the drone onto a predefined mission's path. The experiments are conducted in the simulation software gazebo utilizing a ROS interface to the widely known autopilot software PX4. For obstacle detection, a simulated Intel D435 stereo camera is used. The experiments include two different scenarios, each conducted with two different maximum velocities. The results show that the probabilistic RRT*-algorithm can avoid obstacles successfully and intelligibly even at speeds up to 6 m/s. The main problems persist in the dynamic behavior, the inertia of the multicopter, and the limitations of the sensor technology.