Design and Simulation of an AI-Assisted Autonomous UAV-Surveillance System

Abstract

During the last decade, Unmanned Aerial Vehicles (UAVs) have found their way into different industries ranging from mapping, 3D modeling to thermal inspection. They offer some major benefits over the traditional methods including improvement in the accuracy using high-end sensors, reduction in cost and autonomous operation without any human intervention. Surveillance has also employed drones to monitor larger areas of land from air however generally a human pilot operates or monitors the drone during its flight. In this study, an end-to-end surveillance system is simulated in Gazebo that can provide the object detection functionality using Artificial Intelligence (AI) and can operate independent of any human being. A PID controller is designed and used for precision landing on an ARUCO marker while You Only Look Once (YOLO) algorithm is used to detect and localize objects during the assigned mission. Study indicates encouraging results in detection accuracy and landing precision, suggesting potential for extensive application in different surveillance scenarios. Future work will include real-world deployment and integration with current security infrastructures.