IMU阵列和相机数据融合的定位方法研究

Research on Positioning Method of IMU Array and Camera Data Fusion

在移动机器人领域,通常为机器人配备单个惯性传感器(IMU)和相机来构建视觉惯性SLAM系统,实现定位功能.目前,大多数基于单个IMU和相机所构建的视觉惯性SLAM系统已经能获得良好的定位精度,但在受外界因素影响出现IMU测量异常时,其定位精度会大幅下降甚至导致系统崩溃.对此,本文提出了一种基于IMU阵列和相机数据融合的定位方法.首先基于三次均匀B样条原理获取单个IMU测量值,其次根据IMU阵列信号模型获取IMU阵列测量值,然后利用卡尔曼滤波器对IMU阵列测量值进行融合,得到一个虚拟的IMU数据,提高对IMU故障的鲁棒性.最后为降低位姿估计误差,在OpenVINS框架下融合该虚拟IMU和相机数据进行定位.为验证本文算法的定位精度及鲁棒性,在EUROC数据集和TUM VI数据集上进行了多次实验,实验结果表明在IMU正常或异常情况下,与OpenVINS相比,本文方法能够提供精度更高和鲁棒性更强的位姿估计结果.

In the field of mobile robot,the robot is usually equipped with a single IMU and camera to construct a visual-inertial SLAM system to realize the positioning function.At present,most visual-inertial SLAM systems based on a single IMU and camera have been able to obtain good positioning accuracy.However,when the IMU measurements are abnormal due to external factors,the positioning accuracy will be greatly reduced or even cause the system to crash.In this paper,we propose a positioning method based on the fusion of IMU array and camera data.First,we obtain a single IMU measurements based on the principle of uniform cubic B-splines,and then obtain the IMU array measurements according to the IMU array signal model,and then use the Kalman filter to fuse the IMU array measurements to obtain virtual IMU data,which improves the IMU robustness to failure.Finally,in order to reduce the error of pose estimation,the virtual IMU and camera data are fused under the framework of OpenVINS for positioning.In order to verify the positioning accuracy and robustness of our method,many experiments have been carried out on the EUROC dataset and the TUM VI dataset.The experimental results show that our method can provide more accurate and robust pose estimation results than OpenVINS under normal or abnormal IMU conditions.