Fixed-Wing UAV Autonomous Deployment for Search and Rescue Applications

Abstract

The mission of this project is to design and construct the hardware and software compositions of an unmanned aerial vehicle with fixed wings that has the capability to reliably deploy a small payload to a specified location autonomously. The inspiration and goal of this project is the improvement of efficiency in search and rescue operations. A systems architectural approach was used to produce the final project. Each component of the final drone, sans the deployment system, was purchased off the shelf and integrated into the project framework. The main categories of hardware components include the structural component, flight control, telemetry and communications, manual control, battery power, and motors. As for software components, compatibility with the drone-kit Python library, which provided extensive documentation on its user friendly implementation of drone control, was the most important consideration. A trade study was completed for each component in order to determine the configuration of parts that would provide the lightest, most balanced, and most compatible set up for the drone. Once the drone’s architecture was fully realized, testing of the software and the robustness of the structural design was carried out within the IRoM laboratory in Princeton’s Forrestal campus with the use of the Vicon motion control system. Test of controllability in flight with the weight placement of the final construction was done with manually controlled flight testing. Then, the autonomous deployment code which made use of a data stream of positions from the Vicon motion capture cameras, was tested in several stages. Trials of walking while carrying the drone, named the Disco after its frame that was purchased ready-to-fly initially, was done as an initial measurement. Then, an experimental setup to test the latency or the delay between the dispatch of a command and the physical response was constructed and the results of this test were utilized to improve the autonomous deployment code. Finally, walking, running, and manual flying trials were completed with the improved deployment code. An object detection model for target detection was designed and tested. The model was trained on a custom-built dataset and tested at the Forrestal Campus, but never fully integrated into the Disco’s architecture. The overall result of testing is a drone with a deployment accuracy in flight that is within the constraints of the goal of reaching a person in need.