Title: Quadrotor navigation in GPS-denied environments using multi-sensor data fusion
Language: English
Authors: Natter, Dominik 
Qualification level: Diploma
Keywords: Unmanned Air Vehicle; Multi-sensor data fusion; Visual Odometry; Simultaneous Localization and Mapping; Extended Kalman Filter; Model Predictive Control
Advisor: Kugi, Andreas 
Assisting Advisor: Vu, Minh Nhat 
Issue Date: 2020
Number of Pages: 76
Qualification level: Diploma
The use of unmanned aerial vehicles (UAVs) has increased tremendously over the last decades. With progressing computational power, smaller mechanical setups have evolved, which can enter constrained, GPS-denied environments. This work deals with the development of a state estimator for a quadrotor in such scenarios. On the chosen platform, the algorithms are implemented on an Atom-based board and are partially performed by an autopilot. An open-source flight stack is flashed on the autopilot, taking care of all low-level commands such as motor control. The board runs the Robot Operating System (ROS), which allows to modularly combine different modules. Sensor data from an inertial measurement unit (IMU), a laser-based distance sensor, and an optical flow sensor are available in ROS. Additionally, a Simultaneous Localization and Mapping (SLAM) algorithm determines pose estimates based on images of an RGB-D camera. The proposed state estimator, an extended Kalman filter (EKF) implemented as another C++ ROS node, makes use of this data. It is based on a kinematic model of the quadrotors pose and linear velocity, utilizing quaternions for the rotation representation. A multiplicative formulation of the EKF accounts for the quaternions. The autopilot generates the motor inputs based on the position and orientation setpoints. A further aim of this work is to compare the impact of two different methods for providing the setpoints. On the one hand, a C++ file outputs them one by one after reaching the previous goal. On the other hand, an external time-optimal trajectory planner outputs a number of intermediate points simultaneously. The EKF is tested both in simulation and on real hardware in an office environment. Its performance is worse compared to the one of the autopilots internal state estimator. Still, tracking is possible even during partial loss of measurements. Providing time-optimal trajectories improves the measurement acquisition and is hence the preferable method for delivering the flight commands
URI: https://resolver.obvsg.at/urn:nbn:at:at-ubtuw:1-138680
Library ID: AC15663003
Organisation: E376 - Institut für Automatisierungs- und Regelungstechnik 
Publication Type: Thesis
Appears in Collections:Thesis

Files in this item:

Show full item record

Page view(s)

checked on Apr 11, 2021


checked on Apr 11, 2021

Google ScholarTM


Items in reposiTUm are protected by copyright, with all rights reserved, unless otherwise indicated.