多源融合定位算法综述

A survey of multi-source fusion positioning algorithms

随着机器人、无人车等自主导航系统的大量涌现,定位导航技术在最近20年得到迅猛发展,用户对新一代的定位导航技术提出了新的要求,即在任意环境、任意时刻、任意平台都能具备可靠的定位导航能力.多源融合定位算法是实现该目标的唯一有效途径.本文从传感器观测模型、环境场景模型、载体运动行为模型出发,综述了卫星导航、惯性导航、视觉传感器、激光雷达单一传感器的定位方法,分析了定位导航运行的环境场景对多源融合定位的影响,以及载体运动行为对定位的影响.最后,从融合框架层面将多源融合定位算法分为优化和滤波两大类进行深入分析.

With the emergence of autonomous navigation systems such as robots and unmanned vehicles,positioning and navigation technology has developed rapidly in the past 20 years.However,new user requirements have been raised for update of positioning and navigation,which include reliable positioning and navigation ability in any environment at any time and on any platform.Multi-source fusion positioning algorithm is the only feasible way to achieve this goal.Based on the sensor model,scene model and vehicle dynamics model,we summarize positioning methods of GNSS,inertial navigation,visual sensor,and LiDAR.Then we analyze the influence that the positioning scene and vehicle dynamics pose on the multi-source fusion positioning.Finally,the multi-source fusion positioning algorithms are detailed under two fusion framework categories of optimization and filtering.