A navigation system based on a low-cost, low-grade MEMS inertial measurement unit (IMU) integrated with differential GPS has been developed for machine automation applications. Because the inertial sensors have no ability to measure Earth rotation, the attitude errors of pitch and heading cannot be obtained using only IMU measurements. To overcome this deficiency, two Kalman filters are used for robust estimation of navigation parameters and the errors of inertial sensors. An adaptive fading factor Kalman filter uses a GPS dynamic model to generate the velocities and accelerations which can be used to acquire approximate pitch and heading values. Another Kalman filter is used to integrate position, velocity and attitude from both the IMU and GPS so that position and attitude can be estimated directly - due to their individual observabilities. The drift error of the inertial sensors is also well compensated. The proposed algorithm has been implemented into post-processing integration software and has been tested in the field. The test results demonstrated that this robust MEMS/DGPS integrated system has the capability of providing continuous and reliable navigation for machine automation applications.