Research on ORB-SLAM Autonomous Navigation Algorithm

The autonomous navigation algorithm of ORB-SLAM and its problems were studied and improved in this paper. ORB-SLAM is a fast and accurate navigation algorithm using visual image feature to calculate the position and attitude. But when there are few characteristic points in the unknown environment, ORB-SLAM algorithm falls into the situation of infinite solution, local localization error occurs, and global mapping is interrupted. Researchers have proposed an IMU/Vision fusion algorithm that neatly solves this problem. In this process, the calibration between inertial components and camera is very important, the desirable outcome of which will slow down the divergence speed of autonomous navigation system and further improve the system accuracy. In this paper, the algorithm of calibration between IMU/OE and Stereo vision exploiting the improved recursive least squares with forgetting factor (RFF) method is proposed, and then the algorithm is verified by experimental datas. The results show that the accuracy and the stability of the DR navigation system are improved, which can be applied to IMU/Vision integrated navigation system and enhance robustness of navigation system.