3D Mapping Based IMU Loosely Coupled Model For Autonomous Robot

摘要

In this paper, we focus on building a three-dimensional environment map on large scale scenes for robot navigation. The main contribution of this paper is that we build a loosely-coupled IMU-SLAM fusion model based on EKF and extend it to be a dense mapping system running on embedded devices. To this end, we use the inertial measurement unit to estimate an accurate pose for the direct SLAM system, which improves the efficiency of image matching in direct SLAM and reduces the probability of falling into a local minimum. That is the main reason why we can build dense map robustly and fast on a low-cost platform. We then use the pose of the camera calculated by SLAM to update the IMU bias for the next operating. Finally, the experiments show that our system algorithm can precisely estimate the pose of the camera and robustly build an environment map in large-scale scenes, which can be used in robot navigation.