Improvement of Extended Kalman Filter Using Invariant Extended Kalman Filter

摘要

This paper reports an implementation of invariant extended Kalman filter (IEKF) which improves extended Kalman filter (EKF). The IEKF applies group transformation to state variables and measurement variables based on Lie algebra, thus transforms a non-linear process and measurement system to a locally linear system. Therefore, IEKF approach extends the state space where convergence of states and covariance are guarantees. The implementation estimates the location and attitude of an unmanned aerial vehicle (UAV) using the measurements of the position and velocity from a GNSS, acceleration and angular rate from an IMU, and height by a barometric altimeter. The results show that the estimated location and attitude reasonably tracks the UAV trajectory and the Kalman gain and error covariance converges to constant value.