Visual-based Autonomous Unmanned Aerial Vehicle for Inspection in Indoor Environments

摘要

This paper presents a visual-based autonomous Unmanned Aerial Vehicle (UAV) capable to perform a broad range of navigation tasks in indoor environments. Unlike other related vehicles, our robot does not need an external operator whatsoever or any kind of off-board processing, reducing the overall workload and manual dependency. We evaluated our vehicle in a real indoor environment that simulates an offshore oil platform scenario, demonstrating that it is capable of localize itself, maintaining stable flight, navigating, detecting landing spots, precisely landing on them, and detecting and reading information from the scenario. The methodology presented in this work was used to win the 2019 Brazil Open Flying Robot League during the 2019 Latin America Robotics Competition, among eleven teams and outscoring the second-placing team by almost a factor of two.