Abstract:

For the aim of positioning Extended Kalman Filter (EKF), Unscented Kalman Filter (UKF), and Particle Filter (PF) are used to determine location of moving objects. According to high non-linear model with non-Gaussian noise combining with non-Gaussian noise, the accuracy of EKF becomes worse. To overcome the limitation of EKF, the research focuses on algorithms for non-linear and non-Gaussian including UKF and PF. Root mean square error and computational time are parameters to evaluate these algorithms. In terms of accuracy, PF is the best solution for non-linear model with non-Gaussian noise. The result of PF is more accurate 5 and 9 times than UKF and EKF, respectively. In case of Gaussian noise, the accuracy of UKF is higher 1,5 time than EKF. However, in terms of computational time the EKF is the fastest method while the PF needs a great time to run because of generation of samples.