The Factor Graph Information Fusion Algorithm With Time-varying Noise Estimate for All-Source Navigation System

With the improvement of the performance of the aerial vehicles, the navigation system are needed to adapt to different scenes. The information fusion algorithm, which generate inputs from different sensors and get outputs of the navigation information, is asked to be more adaptive. However, most of current solutions, such as Kalman filters, are aimed at tasks in fixed environments with fixed sensors. The factor graph algorithm based on the principle of probability is a new and valuable method to build the information fusion part of the all-source navigation system. In this paper, the mathematical principles of factor graph is applied to the information fusion algorithm to test its feasibility. Fusion systems based on the local model and global model of factor graph are built, and a time-varying noise estimate is designed to help the systems adapt to changing environments. Their performances are tested by simulations of inputs of different scenes. From simulations, the fusion result is proved to be evidently better than the original inputs, and reach a considerable precision. Moreover, it shows a quick adaptability during the change of working environment or even sensors, which verified its performance. In conclusion, this paper studies the factor graph algorithm and proves its basic feasibility as a new idea of information fusion for navigation system, in order to support the development of all-source navigation system which aims to be functional in different scenes.