Rastreamento de alvos móveis utilizando controle geométrico baseado em visão em veículos aéreos não tripulados

Abstract

This work portrays the development and implementation of a moving target tracking system using vision-based geometric control in unmanned aerial vehicles (UAVs). The main objective was to design a robust controller that would allow UAVs to detect and track targets in real time and autonomous, ensuring precision and stability even in dynamic scenarios. For this, geometric control was used in the SE(3) space, associated with optimization methods such as Particle Swarm Optimization (PSO), aiming to improve the system’s response in comparison to conventional controllers. The research was structured into four main stages. Firstly, a literature review was carried out on geometric control and computer vision applied to UAVs, highlighting existing challenges and gaps. Next, a detailed mathematical modeling of the quadcopter and target was developed, allowing an analysis of the system dynamics and providing a basis for the design of the controllers. Subsequently, a vision-based tracking system was implemented, which enabled the identification and tracking of target UAVs, being validated through simulated experiments in the GAZEBO environment and real ones. Finally, the geometric controller was evaluated, demonstrating superior performance compared to traditional approaches, with lower root mean square error (RMSE) and greater stability in complex maneuvers in the MATLAB simulation environment. The experimental results confirm the effectiveness of the proposed approach, highlighting its applicability in real scenarios.