AUTONOMOUS UAV NAVIGATION: TECHNOLOGIES FOR ORIENTATION AND LOCALIZATION

Abstract

Navigation of autonomous vehicles and particularly unmanned aerial vehicles (UAV) remains one of the core challenges in achieving full autonomy. Autonomous vehicles must independently analyze an environment, react on the dynamic change of surroundings, create optimal trajectories, and perform mission-critical tasks independently. That is why drones need to precisely understand their pose (position and orientation). While in case of error, ground and marine drones might stop the performance, the error of UAV has significant consequences due to the high working speed and heights. It may lead both to losing UAV and harming infrastructure, nature, and civilian houses in accordance to the area, where the UAV is operating.Localization and orientation issues require information from both the state of the drone and information from the environment, as the vehicle’s position in relativeness with other objects is necessary for path planning tasks. This article analyzes the sensors and navigation systems that might give convenient data about both inertial parameters and surroundings. Solutions for both known, partially known, and unknown environments are described. Special attention is paid to Visual SLAM (Simultaneous Localization and Mapping) and Visual-Inertial Odometry, which are key for map creation and relative movement estimation in unknown environments. These approaches enable drones to navigate in GPS-denied or dynamic environments by fusing visual data with inertial measurements.This article also gives a comprehensive review of the map-based sensors, which greatly simplify the UAV performance in pre-known environments. Map-based sensors are applicable for both indoor and outdoor scenarios.The final section of the article introduces the sensor fusion technics, which are necessary for all kinds of UAVs to reduce drifting error and improve navigation accuracy. It enables the comparison of measurements from different sensors or navigation systems with two levels of filtering. Selecting an observer must be done with consideration of the nonlinear nature of autonomous vehicles.