基于可观测度分析和增量因子图的多源融合导航方法

复杂动态场景下高精度导航定位是自主无人系统决策规划与控制执行的基础。在城市峡谷、隧道、地下或室内停车场等场景下,卫星信号弱甚至丢失,严重影响了惯性/卫星/视觉等多源融合导航的精度和可靠性。为主动适应动态场景变化,迫切需要设计一种适用于跨场景的多源融合导航系统。基于动态时变系统的可观测度分析,在线度量惯性/卫星、惯性/视觉等组合导航因子的可信程度,进而采用因子图融合导航框架,根据各组合导航因子的可信程度,主动优化因子构建和增量平滑过程,以实现多传感器自适应融合导航与可靠定位。实验仿真表明基于可观测度分析方法,能够在线优化因子图计算过程,提升了多源融合导航系统的环境适应性和跨场景能力,保证了复杂动态场景下自主无人系统跨场景导航模式切换和连续可靠的导航定位。

Multi-Sensor Fusion Navigation Method Based on Observability Analysis and Incremental Factor Graph

High-precision navigation and positioning in complex dynamic scenarios is the basis for decision-making planning and control implementation of autonomous unmanned vehicles. In general, satellite signals in urban canyons, tunnels, underground or indoor parking lots are weak or even lost, which greatly affects the accuracy and reliability of INS/GNSS/Vision multi-sensor fusion navigation. In order to actively adapt to dynamic scenario changes, it is urgent to design a multi-sensor fusion navigation system that is suitable for cross-scenario applications. Based on the observability analysis of dynamic time-varying system, the confidence level of INS/GNSS or INS/Vision integrated navigation factors can be evaluated online. Furthermore, according to the degree of confidence of each integrated navigation factor, a fusion navigation framework of incremental factor graph is adopted to optimize the factor construction and incremental smoothing process, so as to achieve multi-sensor adaptive fusion navigation and reliable positioning. The simulation results show that the factor graph can be optimized online based on the observability analysis method, which improves the environmental adaptability and cross-scenario capability of the multi-sensor integrated navigation system, enabling the adaptive switch of cross-scenario navigation as well as continuous and reliable positioning of autonomous unmanned vehicles in complex dynamic scenarios.