Host: The Japanese Society for Artificial Intelligence
Name : The 32nd Annual Conference of the Japanese Society for Artificial Intelligence, 2018
Number : 32
Location : [in Japanese]
Date : June 05, 2018 - June 08, 2018
This paper describes a method for estimating the flight area of drones based on deep learning. The position of a drone can be detected by using the global positioning system (GPS). However, GPS sometimes has problems capturing signals from satellites that are shielded by mountains and/or buildings. Moreover, GPS signals are very weak and subject to a variety of disturbances. Such problems increase the chance of a crash when flying a GPS controlled drone. As a solution to this problem, we propose a flight area estimation method using a 3D map created on the basis of deep learning. Our method could estimate the flight area with 98.4 percent accuracy in a field experiment.