Obstacle voidance for Unmanned Aerial Vehicles during teleoperation
Date published
Free to read from
Authors
Supervisor/s
Journal Title
Journal ISSN
Volume Title
Publisher
Department
Type
ISSN
Format
Citation
Abstract
Unmanned Aerial Vehicles (UAVs) use is on the rise, both for civilian and military applications. Autonomous UAV navigation is an active research topic, but human operators still provide a flexibility that currently matches or outperforms computers controlled aerial vehicles. For this reason, the remote control of a UAV by a human operator, or teleoperation, is an important subject of study. The challenge for UAV teleoperation comes from the loss of sensory information available for the operator who has to rely on onboard sensors to perceive the environment and the state of the UAV. Navigation in cluttered environment or small spaces is especially hard and demanding. A flight assistance framework could then bring significant benefits to the operator. In this thesis, an intelligent flight assistance framework for the teleoperation of rotary wings UAVs in small spaces is designed. A 3D haptic device serves as a remote control to improve ease of UAV manipulation for the operator. Moreover, the designed system provides benefits regarding three essential criteria: safety of the UAV, efficiency of the teleoperation and workload of the operator. In order to leverage the use of a 3D haptic controller, the initial obstacle avoidance algorithm proposed in this thesis is based on haptic feedback, where the feedback repels the UAV away from obstacles. This method is tested by human subjects, showing safety benefits but no manoeuvrability improvements. In order to improve on those criteria, the perception of the environment is studied using Light Detection And Ranging (LIDAR) and stereo cameras sensors data. The result of this led to the development of a mobile map of the obstacles surrounding the UAV using the LIDAR in addition to the stereo camera adopted to improve density. This map allows the creation of a flight assistance system that analyses and corrects the user’s inputs so that collisions are avoided while improving manoeuvrability. The proposed flight assistance system is validated through experiments involving untrained human subjects in a synthetically simulated environment. The results show that the proposed flight assistance system sharply reduces the number of collisions, the time required to complete the navigation task and the workload of the participants