Unmanned Aerial Vehicle (UAV) development is a field with rising interest in recent years. One of its essential part is the autopilot, responsible for keeping the aircraft in desired flight conditions and for executing navigation tasks. Among the navigation tasks, one that stands out is the pathfollowing operation, which guarantees that the aircraft follows a predefined trajectory. The literature presents a variety of approaches with this function, based in geometric and control techniques; however, compared only for the 2D scenario. Therefore, the main purpose of this paper is to perform a comparison in the 3D scenario of the path-following algorithms Lookahead, Non-Linear Guidance Law (NLGL), Pure Pursuit and Line-of-Sight (PLOS) and Vector Field, regarding accuracy and computational cost. The simulation results of straight lines trajectories show that PLOS and Lookahead present lower errors and lower computational cost than NLGL and Vector Field, thus being more adequate.

The Extended Kalman Filter (EKF) and Unscented Kalman Filter (UKF) are methods usually applied in the sensor fusion for Unmanned Aerial Vehicles due to its nonlinear navigation equations. This paper presents a comparison between the two filters considering the position, velocity and attitude of the vehicle and the IMU bias. The simulation experiments are designed according to performance evaluation techniques for two trajectories and different state vectors. The results show that the EKF has a lower computational cost than UKF, but the latter provides smaller errors for most of the states. It also show that the bias estimation influences positively the solution granted by the EKF.