Abstract
This paper develops a method that uses two magnetometers to detect whether the indoor magnetic field is disturbed for mobile robot localization purposes. We define a confidence index to show the degree of interference. The relationship between confidence index and variance of the orientation error derived from magnetometers is established by numerical simulations. Then this relationship is applied to perform the measurement update in Kalman filter to correct the orientation prediction information from wheel encoder. Thus we can get more accurate orientation information from magnetometers under magnetic interference. Then we use this orientation information to correct the error from prediction model which contains the gyroscope as input. Thus we obtain the orientation that has good transient performance due to the use of the gyroscope and good steady-state performance from magnetometers. Finally, we use the extended Kalman filter (EKF) localization to combine multi-sensor data and image processing information from camera to correct the odometry error of mobile robots. Experimental results verify that the proposed method can achieve satisfactory localization performance.