Controle preditivo baseado em dados para rastreamento de trajetórias em quadricópteros

Abstract

This study investigates the application of the Data-Enabled Predictive Control (DeePC) algorithm to the trajectory tracking of multirotor unmanned aerial vehicles. The objective of this work is to demonstrate the feasibility of predictive control based exclusively on input-output data, bypassing the parametric identification of mathematical models of the system. Drawing from Willems' Fundamental Lemma, which establishes that trajectories of linear time-invariant systems can be represented as linear combinations of persistently exciting historical data, a regularized formulation is developed to operate under conditions of noise and moderate nonlinearities. For validation, a high-fidelity simulation environment is implemented based on a reference quadcopter, incorporating nonlinear dynamics, actuator saturation, and computational delays. Systematic comparisons are conducted with model-based controllers, specifically PID optimized via Particle Swarm Optimization (PSO), the Linear-Quadratic Regulator (LQR), and Model Predictive Control (MPC). The results demonstrate that the data-driven approach achieves tracking performance comparable to classical predictive control, significantly outperforming reactive strategies. Regarding robustness, the analysis reveals that the controller maintains acceptable performance even in the presence of measurement noise, establishing itself as a promising alternative for applications where precise mathematical modeling is impractical, offering a balance between precision, robustness, and implementation simplicity.