Abstract:Map match is a mainstream on board positioning method. Based on the track information and road map, the vehicle position is estimated on the road map and positioning error can be updated by this method. And this method becomes the base of vehicle navigation, traffic guidance and traffic prediction. But there are disadvantages of current map match algorithms, such as match process relying on road searching precision and road searching fail with large initial position error which cause following match error. A map match algorithm based on graphic convolution is presented to avoid the error caused by incorrect intersection match and large initial error. The map match method algorithm based on graphic convolution contrast the geometry characteristic of vehicle track and road to locate vehicle position. This map match algorithm is fit in large initial position error scene and avoid road searching fail problem.