Proposal of Autonomous Playback Navigation System for an Okinawa-Style Drone

Abstract

The purpose of this research was to develop the basic concepts needed for an experimental ‘human-friendly’ drone that can enable flexible flight and support tourism. The developed concept is to design a drone capable of autonomous transfer and intelligent operation by performing operations as instructed in response to a call from an operator. The drone learns a safe path by following human instructions and then plays back the learned information while autonomously flying along the learned path. We call this system “playback navigation”.  While systems for mobile robots have used odometry for robot position detection, we used the new “Air-odometry”, which detects the drone's position by calculating it from acceleration sensors. Air-odometry is also used by the drone to learns a safe path by following a human instructions. A linear approximation of the result is calculated and used as the path during autonomous playback navigation.  To enable autonomous navigation around corners of the learned path for which even SLAM algorithms tend to generate errors, we placed a camera-driven position correction landmark at each corner. We were able to get good results.