Development of low-cost camera-based UAV control strategy

Authors:

Abstract
Nowadays, UAV can fly well with GPS outdoors. However, when there is no GPS, it is hard to control the UAV. Our control method aims at indoor UAV control, where there is no GPS. We use a down-looking camera and imaging processing approach to control the UAV. Through our method, the UAV can depart and land in the specific area as well as following the path set on the ground. We have our method proven by a simulation environment. For obstacle avoidance task, a front-looking camera and Dronet is employed. Dronet can predict the collision probability. Whenever the probability is too high, the UAV would hover till the obstacle, such as a human, disappears from its vision. Then we compress Dronet to make it perform faster without too much loss of accuracy on embedded computers.
Funding Information
  • MOST Foundation (MOST 110-2221-E-027-118)

This publication has 13 references indexed in Scilit: