| Resumo: | With the increasing sophistication of smartphone applications, accurate positioning and tracking have become a vital component of various cutting-edge technologies where sensor fusion is essential in the integration of data from multiple sensors to achieve a more accurate and reliable estimation of the state of a system. This project proposes an innovative approach to improve the performance of Kalman filter algorithms, specifically, the Enhanced Error-State Kalman Filter (ESKF), for smartphone pose estimation. The ESKF method is a more efficient and robust alternative to the commonly used Extended Kalman Filter (EKF) method, particularly for nonlinear systems. ESKF requires prior construction of an error process that relates the error variables associated with the pose estimation system model. Besides, a magnetometer correction model is introduced which enables better magnetic tolerance and compensation than the conventional IMU filter. The results demonstrate that the enhanced ESKF method achieves remarkable precision in estimating three-dimensional rotations with error typically below 5° under a non-disturbed environment.
|
|---|