Multiple Fading Factors Extended Kalman Filter Based on Mahalanobis Distance for Robot Localization Under Measurement Abnormality

This paper investigates the problem of mobile robot positional estimation with robot localization under measurement abnormality. In order to improve the accuracy of position estimation without increasing the computational complexity, multiple fading factors extended Kalman filter based on Mahalanobis distance is proposed. The criterion of measurement abnormality is constructed based on the Mahalanobis distance of the innovation vector and the hypothesis testing method. A multiple fading factors matrix is constructed to provide different fading rates for different measurement data channels by combining the scalar fading factor and the innovation covariance, which realizes the adaptive adjustment of the statistical characteristics of measurement noise. Simulation results under two typical measurement abnormality scenarios verify the effectiveness and robustness of the proposed algorithm.