基于视觉-IMU传感器融合SALM研究

Research on SALM based on vision⁃IMU sensor fusion

为了验证提出的多传感器融合在工程应用中的有效性,针对视觉运动产生的失真情况,提出一种视觉-IMU传感器融合SALM的研究,并且设计了多传感器融合SLAM系统,将相机与IMU进行联合标定、采用图优化的融合方法在空间和时间上得到数据上的对齐。通过ROS系统自带的GAZEBO进行实时仿真,建立机器人小车模型在仿真的空间中运动,再通过APE绝对位姿误差计算并对比,得出了视觉-IMU融合的绝对位姿误差要比未融合情况下的误差小,优化之后的精度相比未融合的精度要高。因此利用该方法设计的融合系统较为合理。

In order to verify the effectiveness of the proposed multi⁃sensor fusion in engineering applications,aiming at the distortion caused by visual motion,a vision⁃IMU sensor fusion SALM is proposed,and a multi⁃sensor fusion SLAM system is de⁃signed.Perform joint calibration and use graph⁃optimized fusion methods to obtain data alignment in space and time.Real⁃time simulation is carried out through GAZEBO that comes with the ROS system,and the robot car model is established to move in the simulated space,and then the absolute pose error of APE is calculated and compared,and the absolute pose error of vision⁃IMU fu⁃sion is obtained.The error is small,and the accuracy after optimization is higher than that without fusion.Therefore,the fusion sys⁃tem designed by this method is more reasonable.