Improving iterated Extended Kalman Filter for non-Gaussian noise environments

Kalman filter is an efficient algorithm to estimate the state of linear time-discrete dynamical systems with Gaussian noises. Unfortunately, practical systems are mostly non-linear and noise follows non-Gaussian distributions. This article proposes an improvement in Extended Kalman Filter (EKF) algorithm for predicting non-Gaussian distributed state noise. Further introducing the proposed scheme into the indoor positioning system illustrates that considerably higher accuracy in locating and tracking objects can be achieved.