单目视觉惯性SLAM与UWB数据融合的精确定位

Precise localization of monocular vision inertial SLAM and UWB data fusion

针对室内或封闭场景下的无人驾驶,提出了一种融合视觉同步定位与建图(SLAM)与超宽带(UWB)的定位方法。通过惯性测量单元(IMU)预积分提高单目SLAM的可靠性;再将视觉SLAM与UWB获取的定位信息通过扩展卡尔曼滤波器(EKF)得到融合后的数据。实验结果表明:在正常光照下环境下的最大误差为0.101 316 m,根均方误差(RMSE)为0.035 965,在弱光照环境下的最大误差为0.136 103 m, RMSE为0.039 132,未出现明显的跳变及漂移。提高了系统定位精度及鲁棒性。

Aiming at unmanned driving in indoor or closed scenes, a positioning method which fuses visual simultaneous localization and mapping(SLAM)and ultra-wideband(UWB)is proposed.The reliability of single SLAM is improved by inertial measurement unit(IMU)pre-integral.Then, the position information obtained by visual SLAM and UWB is fused with the extended Kalman filter(EKF).The results show that the maximum error under normal light is 0.101 316 m, root mean square error(RMSE)is 0.035 965,maximum error under weak light is 0.136 103 m, RMSE is 0.039 132,and no obvious jump or drift occurs.The positioning precision and robustness of the system are improved.