© 2017 IEEE. In our previous study, we developed the odometer having a caster mechanism. We can estimate the translational and rotational velocity of a mobile robot using two of them. When we use the odometers on an uneven ground, the measured values of the odometers change rapidly, which renders difficulty on accurate estimation of vehicle velocity. In this study, we calculate Mahalanobis distance by the algorithm of Extended Kalman Filter (EKF). Using the weight calculated by Mahalanobis distance, we estimate the velocities of the robot equipped with the odometers by Moving Horizon Estimation (MHE). In this paper, we show the effectiveness of the proposed method under the uneven ground by simulation.