A fuzzy extended Kalman filter for mobile robot localization in high amount of measurement noise

Localization is a crucial part in robot navigation. The significance of this issue is to the extent that further achievements in terms of robot control is highly dependent on that. There are various methods addressing this subject, among which the extended Kalman filter is proved to be one of the most successful ones. In classic extended Kalman filter, covariance matrixes of process noise Q and measurement noise R in the associated cost function consist of fixed elements. Therefore, in case of uncertainties or noises, the algorithm has some inherent limits to show sufficient flexibility to conform to new conditions. In current approach a new Fuzzy extended Kalman filter is proposed in which the matrix Q is considered as a Mamdani-type fuzzy inference system. In the proposed fuzzy system, fuzzy rules are carefully designed such that the new hybrid scheme significantly outperforms the classic extended Kalman filter in high amount of measurement noise particularly when the noise interference is considered.