1.2085m减小到0.4861m。Since the positioning accuracy of low accuracy GPS can not meet the requirements of agricultural production, a new navigation method is studied to improve the position accuracy. The compensate model is built to compensate for the libration caused by the antenna. A Kalman filter is developed, and a VRS-based receiver is used as the low accuracy GPS. The Kalman filter is applied to the data collected from the low accuracy GPS, compass and speed sensor. The data set collected from the RTK-GPS is used as a baseline to evaluate the performance of the filter. Experiment results indicated that the mean error was reduced from 1.6019m to 0.597m straight line track, and from 1.2085 m to 0.4861 m in curve track.