Rejection of drift disturbances in robot navigation

In this paper the problem of robot navigation based on GPS and odometry is considered. One of the common problems with odometry systems is the presence of disturbances associated with the measure. A new method based on adaptive Kalman filtering is presented for these cases. The method is designed for those cases in which the process disturbance affecting the system is white noise but nonzero mean. This causes a drift in the estimation result of the filter. The technique proposed is evaluated in simulation and compared with other established methods.