Autonomous navigation for flying quadcopters using VSLAM

Abstract

When it comes to moving robots, there is a major drawback in autonomous navigation, as adding the sense of sight to a machine can be very complicated. There are other sensors that can help a robot interact with the environment, such as ultrasonic distance sensors, used to prevent collisions with objects or pressure sensors, to know the height of the robot. In addition, GPS navigation methods can be very useful for managing robot movements in an open space more or less accurately. However, for indoor applications where the GPS signal is deficient or non-existent, we must fully rely on the data from our sensors. A sensor that, for this project, will be a 2D camera without any depth information. In this paper we will study the LSD-SLAM System with integration with ROS technology. All tests will be performed using the Gazebo simulator, in a virtual space. All procedures followed in this project have been tested on Ubuntu 16.04. This document describes all the steps to install and configure ROS technology, along with the aforementioned simulator and the LSD-SLAM system. In addition, flight tests will be carried out indoors and outdoors where we will check the effectiveness of the algorithm. This will be quantified and measured between the two scenarios. All the packages needed to be able to rebuild the project can be found in the references.