基于联邦滤波的室内多传感器融合导航定位方法研究

Indoor Multi-Sensor Fusion Navigation and Positioning Method Based on Federated Filtering

针对室内多传感器融合导航定位问题,提出一种基于联邦滤波的INS/3D LASER/单目视觉融合的室内导航定位方法。该方法是通过联邦滤波将INS分别同3D LASER与单目视觉构成子滤波器,主滤波器根据最小方差准则,利用最优系数加权法融合各子滤波器的最优估计,从而获得全局最优融合结果。实验证明,该方法能有效提高室内导航定位的精度和鲁棒性,比传统的视觉里程计法或激光雷达里程表法更准确。

To solve the problem of indoor multi-sensor fusion navigation and positioning,an indoor navigation and positioning method based on INS/3 D LASER/monocular vision fusion is proposed.In this method,INS is combined with 3 D LASER and monocular vision respectively to form sub filters through federated filtering.According to the minimum variance criterion,the main filter uses the optimal coefficient weighting method to fuse the optimal estimation of each sub filter,so as to obtain the global optimal fusion result.Experiments show that this method can effectively improve the accuracy and robustness of indoor navigation and positioning,which is more accurate than the traditional visual odometer method or the LiDAR odometer method.