基于多源定位的机器人导航地图融合研究

Research on robot navigation map fusion based on multi-source localization

针对激光雷达(LiDAR)的传统建图方式在精度和地图完整性上存在不足的问题,提出一种以多源定位数据实现创建融合栅格地图的方法。利用误差状态卡尔曼滤波(ESKF)算法将惯性测量单元(IMU)与轮式编码器(WE)的融合数据中添加视觉里程计(VO)的位姿信息进行校正,并作为里程计输出。根据贝叶斯估计,将深度相机与激光雷达各自生成的局部栅格地图逐帧进行融合,生成全局地图。研究结果表明:融合地图与实际环境中的对应参考点在x与y方向的RMSE比传统方法分别下降了58.88%,56.19%,有效提高了地图的精度和丰富性。

Aiming at the problems of precision and map integrity in traditional map construction method of LiDAR,a method to create fused raster map based on multi-source localizaion data is proposed.The pose information of visual odometer(VO)is added to the fusion data of inertial measurement unit(IMU)and wheel encoder(WE)by using error state Kalman filtering(ESKF)algorithm,which is used as odometer output.According to Bayesian estimation,the local grid maps generated by depth camera and LiDAR respectively are fused frame by frame to generate a global map.Research results show that the RMSE of the corresponding reference points in the x and y directions of the fusion map and the actual environment decreases by 58.88%and 56.19%,respectively,compared with the traditional method,which improves the precision and richness of the map effectively.