Nombre: DIEGO ABRAHAM APAZA LAURA
Fecha de publicación: 17/12/2025
Junta de examinadores:
Nombre |
Papel |
|---|---|
| DANIEL KHEDE DOURADO VILLA | Presidente |
| JOSE LEANDRO FELIX SALLES | Examinador Interno |
| KEVIN BRAATHEN DE CARVALHO | Examinador Externo |
| WANDERLEY CARDOSO CELESTE | Examinador Interno |
Sumario: This dissertation presents the design, implementation, and experimental validation of a Model Predictive Control (MPC) framework for trajectory tracking and obstacle avoidance in Unmanned Aerial Vehicles (UAVs) of the quadrotor type. The strategy focuses on safe navigation in dynamic environments. The core of the approach consists of an avoidance mechanism that linearizes the non-convex collision constraints at each step of the prediction horizon using time-varying tangent planes. This allows the optimization problem to be efficiently formulated and solved as a standard Quadratic Program (QP). To enhance robustness and ensure solver feasibility in highly constrained scenarios or in the presence of disturbances, the optimization problem incorporates flexible constraints. This technique uses slack variables to allow temporary violations of safety zones at a high penalty cost, ensuring a feasible solution is always found. The controller was implemented in MATLAB and executed on a ground station, communicating with the UAV in real-time (30 Hz) via the Robot Operating System (ROS). Validation experiments were conducted using a real quadrotor (Parrot Bebop 2), whose pose was provided by a motion capture system (OptiTrack). Test scenarios included trajectory tracking and setpoint regulation in the presence of multiple static and dynamic obstacles (Pioneer 3-DX ground robots). The results demonstrate that the proposed system enables the UAV to navigate efficiently and smoothly, anticipating and executing safe evasive maneuvers. The successful completion of all experimental scenarios validates the effectiveness and reliability of the approach for safe operation in complex environments.
