基于MCMCC改进ORB-SLAM3的移动机器人导航

Improved ORB-SLAM3 Based on MCMCC for Mobile Robot Navigation

【目的】针对VSLAM建立的稀疏特征地图不能直接用于导航,且现有的将基于稀疏特征图转化为栅格地图进行导航的方法存在建图精度不高的问题,无法实现准确导航。提出了一种基于MCMCC改进ORB-SLAM3的移动机器人导航实现方法,在仅有深度相机的嵌入式设备上可以实现室内实时地图构建、定位与导航。【方法】提出的方法利用基于多凸组合最大互相关熵(MCMCC)标准的后端优化算法来改进ORB-SLAM3的后端优化模块,降低非高斯噪声对建图精度的影响。同时在将稀疏特征图转化为栅格地图的过程中,利用回环信息更新栅格地图,以提高栅格地图的精度。【结果】仿真和真实场景实验验证了所提出方法的有效性。

【Purposes】 The sparse feature map established by using VSLAM cannot be directly used for navigation, and the existing methods of converting sparse feature maps into grid maps for navigation suffer from low accuracy of map construction, resulting in the failure to achieve accurate navigation. An improved ORB-SLAM3 navigation method is proposed based on multi-convex combined maximum correntropy criterion(MCMCC), which can realize the construction, positioning, and navigation of indoor real-time mapping on the embedded device with only a RGB-D camera. 【Methods】 The proposed method uses the back-end optimization algorithm based on MCMCC to combat the influence of non-Gaussian noises and to further improve the mapping accuracy of ORB-SLAM. Subsequently, in the process of converting the sparse feature map into a grid map, the grid map is updated with the loop detection information to improve the grid map accuracy for accurate navigation. 【Findings】 Furthermore, simulation and real scene experiments have been carried out to verify the effectiveness of the proposed method.