Using autonomous drone navigation to pick up and deliver payloads

Abstract

This project aims at researching and investigating how to employ unmanned aerial vehicles (UAV) efficiently, also referred to as drones, to optimally pick up and deliver a package from source to destination entirely autonomously. The primary motivation for this project came from watching Amazon’s drone delivery service at work; interest in the subject grew. In knowing that a full-scale model was not achievable due to all the constraints, a workaround needed to be adopted. To cater for this, a small-scale model was to be developed and used extensively to research the subject. An off-the-shelf drone and Lego Bricks were used to their best ability to try and mimic what would have been achieved using the full-scale model. The drone being used is a Tello EDU, this is a very small programmable drone only measuring up to 10cm in length. Given this size, the drone is very lightweight and considerably fast. Although its main drawback is its battery life which averages out at around 13 minutes. Having only access to such a small drone, meant that the entire project had to be tailored in such a way that no tests exceed its specified capabilities. A dataset filled with various pictures of payload made out of Lego Bricks, fitted with a magnet, was created. This dataset was to be used later for the training of an object detection algorithm using YOLOv3. The trained object, later on, shifted to act as a target which would prompt the drone to descend and mimic picking up a payload. Being a proof-of-concept, the entire model containing autonomous flight and object detection was then tested, and appropriate results were achieved.