Autonomous vehicle navigation with standard IMU and differential GPS has been widely used for aviation and military applications. Our research interesting is focused on using some low-cost off-the-shelf sensors, such as strap-down IMU, inexpensive single GPS receiver. In this paper, we present an autonomous vehicle navigation method by integrating the measurements of IMU, GPS, and digital compass. Two steps are adopted to overcome the low precision of the sensors. The first is to establish sophisticated dynamics models which consider Earth self rotation, measurement bias, and system noise. The second is to use a sigma Kalman filter for the system state estimation, which has higher accuracy compared with the extended Kalman filter. The method was evaluated by experimenting on a land vehicle equipped with IMU, GPS, and digital compass.