Bayesian Optimization for Fine-Tuning EKF Parameters in UAV Attitude and Heading Reference System Estimation
Published in Aerospace, 2023
Accurately estimating a vehicles 3D motion and seamlessly fusing data from multiple sensors are crucial tasks in various applications. While the Extended Kalman Filter (EKF) reigns supreme in this domain, its effectiveness hinges on optimal tuning of two key parameters: the process and measurement noise covariance matrices (Q and R). Choosing these values correctly can be a daunting challenge, hindering the EKFs full potential. This research tackles this challenge head-on by proposing an EKF innovation consistency statistics-driven Bayesian optimization algorithm. This novel approach automatically tunes Q and R based on the desired performance criteria, specifically, minimizing estimation error through improved measurement innovation consistency. Our extensive results showcase a significant performance boost in the EKF when equipped with the optimized Q and R values obtained through our algorithm. In essence, this research paves the way for easier and more effective utilization of the EKF in various vehicle motion estimation and sensor fusion applications.
Recommended citation: Wondosen, A.; Debele, Y.; Kim, S.-K.; Shi, H.-Y.; Endale, B.; Kang, B.-S. Bayesian Optimization for Fine-Tuning EKF Parameters in UAV Attitude and Heading Reference System Estimation. Aerospace 2023, 10, 1023. https://doi.org/10.3390/aerospace10121023 https://doi.org/10.3390/aerospace10121023