Getting Started with a Simple Visual-Inertial Odometry

This paper presents the development and evaluation of a basic monocular visual-inertial odometry (VIO) algorithm tailored for drone navigation. The primary goal is to present a minimal, easy-to-understand algorithm for beginners to the field. Therefore, the method utilises proven, simple, non-computationally intensive computer vision algorithms. To solve the unknown scale problem, monocular visual odometry estimates are fused with scale estimates extracted from inertial measurement unit (IMU) data. We tested the algorithm on various datasets, including the real-world KITTI dataset and a custom dataset recorded with a Tello DJI drone in a controlled lab environment. Moreover, the algorithm's adaptability and performance were further validated in an Unreal-Engine simulated environment utilizing Microsoft AirSim for the simulation of the drone's physics. Although simple, the proposed method is versatile for short-term odometry especially for slow continuous movements.