Attitude and Heading Estimation of UAV Using Quaternion and Particle Filter

Abstract

Accurate estimation of attitude and heading in UAVs is one of the key challenges in autonomous navigation systems, playing a vital role in the control and guidance of these vehicles. In this paper, a quaternion-based particle filter with a specific sampling method and an Uncented Kalman filter (UKF) are employed for UAV attitude estimation. By integrating data from inertial sensors (including gyroscopes, accelerometers, and magnetometers) and applying filtering techniques, the proposed methods significantly enhance the accuracy of roll, pitch, and yaw angle estimation. The innovation of this study lies in the comprehensive comparison of the particle filter and UKF performance across two scenarios: simulated data and real-world data collected from a specific attitude and heading reference system (AHRS). The results demonstrate that the particle filter achieves remarkable improvements of over 99.99% in simulated data and over 88.48% in real-world data for roll and yaw angle estimation. Additionally, the analysis of variance and standard deviation of errors confirms that the particle filter outperforms in reducing error dispersion, with the error variance for yaw angle being approximately 100 times lower than that of the UKF. On the other hand, the UKF shows slightly better performance in pitch angle estimation. These findings suggest that a combination of these two filters can serve as an effective solution for precise navigation systems in UAVs. The resampling method implemented in the particle filter also significantly enhances the accuracy of roll and yaw angle estimation.