UAV System Integration of Real-time Sensing and Flight Task Control for Autonomous Building Inspection Task

摘要

Most current research on autonomous UAV control are conducted in virtual environment and often focus on specific technical domains, such as communication protocol and visual computing, etc. These research are rarely integrated as one, making it less applicable and effective. However, application platforms solving practical problems require system integration in real-world environments. In this paper, we propose a UAV task-oriented flight control system integrated with real-time sensing based on Robot Operating System. The flight control system also incorporates behavior tree as a decision control mechanism. We apply our system in a building inspection task, whereas the UAV takes-off near the riverbank, flies forward following the road across a bridge, and arrives at a designated building to perform a zig-zag image scan. The system is implemented and field tested on a ready-to-fly quadrotor. Captured images are transmitted through Wi-Fi to a laptop for real-time visual sensing, and flight directions are then calculated and provided to the UAV for navigation. Our experiment shows promising results - the UAV can successfully complete the task within 25 minutes. Some suggestions and solutions are provided for future improvements.