多源融合导航系统的因子图算法综述

A Survey of Factor Graph Algorithms for Multi-source Fusion Navigation System

由于卫星导航系统不能穿透物理遮蔽物且容易受到干扰以及惯性导航系统误差积累容易发散的特点,传统的惯性/卫星组合导航系统越来越难以满足人们对导航精度的需求,于是多传感器组合导航系统应运而生并不断发展。在多源导航系统中,如何高效可靠地融合不同传感器的数据是其中的核心,工程实践中采用较多的是联邦Kalman滤波,但其在处理异步异质非周期信号和非线性问题时都显得力不从心,近来越来越多的目光聚集在因子图算法上,故对目前因子图在多源组合导航系统的应用情况以及将来可以发展的方向进行梳理阐述。

Due to the fact that satellite navigation systems can not penetrate physical obstructions and are susceptible to be interfered, and the accumulation of errors in inertial navigation systems are prone to diverge, traditional inertial/satellite integrated navigation systems are increasingly difficult to meet the needs of navigation accuracy, so multi-sensor integrated navigation system arises at the historic moment and develops continuously. In multi-source navigation system, how to fuse the data of different sensors efficiently and reliably is the core of the system. In engineering practice, the federated Kalman filtering is the most common method. However, it is not able to handle asynchronous heterogeneous aperiodic signals and nonlinear problems. Recently, a increasing more attention has been focused on the factor graph algorithm. Therefore, the application of factor graph in multi-source integrated navigation system and the future development direction are summarized.