Optimized Unmanned Aerial Vehicle (UAV) Localization and Autonomous Navigation Stack for Tightly Closed Industrial Spaces
Abstract: Inspection automation of industrial infrastructure and surveillance became an integrated part of the new industrial environments in the age of the fourth industrial revolution and digital transformation; thus, the need to remove manual labor from such processes to improve efficiency and reduce costs became necessary. Recent advances in sensor technology, edge devices, and unmanned aerial vehicle (UAV) control systems have made it possible to develop systems with such capabilities using off-the-shelf components and integrated platforms. This research work aims to identify and develop a navigation stack for localization and navigation in the absence of global positioning system (GPS) signals. Furthermore, industrial environments are frequently complex in terms of navigation space; thus, such systems must have a high precision rate for pose estimation. Tanks and pipes for storage and transport, which require inspections for safety and are usually costly and risky, are also common in an industrial environment. As a result, UAVs can offer a safe and profitable alternative solution in addition to precision and speed. A working prototype is developed for a navigation system that enables autonomous navigation for a UAV in a simulated industrial environment using a local path planning module with estimated pose data from a localization module using visual odometry sensor data. The results show an average error rate of 10-30 cm, which is promising for navigation within complex spaces with a range of 1.5-8.5 m. The developed stack can serve as the foundation for future development to improve pose estimation precision and performance while also increasing the stability of the local path planning algorithm. Furthermore, the developed system represents a one-way communication channel between the localization module and the local path planning module. However, bi-directional communication to provide feedback from the local path planning module to the localization module can improve precision.
Loading