Modern mobile robots often include multiple sensors that measure correlated physical quantities. The use of an extended Kalman filter allows data to be fused in a way that the result is more reliable than any individual sensor measurement. On a differential driven mobile robot equipped with an inertial measuring unit and wheel encoders, we demonstrated the use of EKF in order to improve reliability of the robot's tracking. Tested robot is based on ROS firmware, which allowed us to use the open source robot\_localization software package in which this filter is already implemented. The EKF tracking test was performed by driving the robot along a track and comparing the results to the position acquired from the external camera. We have found that data from the inertial measurement unit is not reliable enough to improve on the robots localization. Nevertheless, we concluded that filtering data from encoders alone increases the reliability of localization, as the filter continues to predict the status of the robot, even when sensor data is temporarily unaccessible.