Research On Indoor 3D SLAM Based On 2D Rotating Laser

With the continuous development of Virtual Reality(VR)and Augmented Reality(AR)technologies,people's demand for surrounding environment perception,positioning and modeling is increasing.At the same time,indoor positioning and navigation has gradually developed into a 3D model map.Therefore,it is particularly important to accurately locate in indoor scenes and construct environment maps conveniently,quickly and at low cost in the current context.Simultaneous localization and mapping(SLAM),as a basic module in many fields,can determine its own position in real time,and at the same time build a map of the surrounding environment.It has the characteristics of rapid and flexible data collection,real-time response,portable and easy operation of equipment,and becomes one of the current research hotspots.Aiming at the problems that 2D laser SLAM can't meet daily demands,traditional 3D laser SLAM cost is high,and the existing methods of indoor scene positioning and composition are not robust,this paper designs and implements a 3D laser SLAM system using 2D laser sensors for indoor data collection,It has the characteristics of low cost,easy to carry,and expandable.It can quickly build an indoor environment map and determine its position in real time.The hardware acquisition system only relies on the motor to drive a 2D laser sensor to rotate around the forward axis,and scans to obtain a 3D point cloud of the scene.The structure is simple and the cost is low.Considering the characteristics of indoor structure,SLAM Algorithm in this paper is based on the widely existing plane and plane intersection line(Intersection line)as the high differentiation feature,and uses robust feature matching to carry out feature association.Through rough estimation of single frame deformation,fast frame linear optimization,fine map B-spline optimization,the hierarchical pose optimization strategy balances efficiency and precision,and gradually optimizes the pose trajectory to solve the problems of low-cost acquisition equipment,such as small amount of data,low frequency,large frame deformation,etc.,to achieve a low-cost,low drift indoor laser slam system.In addition,for the feature expression with both local and global information,this paper also proposes a structured scene map representation method constructed by plane and plane intersection features and their adjacency relationships,which can concisely and effectively express the indoor scene skeleton structure.The strong structure and semantic information can facilitate the rapid feature association and map update,and is suitable for indoor scene expression.In the experimental part of this paper,qualitative and quantitative experiments are carried out to verify and evaluate the algorithm.First of all,through the ablation experiments of different modules of the algorithm,the effectiveness of the high differentiation feature expression and pose optimization from coarse to subdivision level proposed in this paper is verified;then,from the aspects of hardware time delay error,delay noise,inaccurate attitude of motor angle,data loss and other experiments,it is verified that the method in this paper has low requirements for hardware design and installation,and does not need to be calibrated in advance;at the same time,from the stability point of view The robust feature expression matching and hierarchical pose optimization strategy are compared to verify that this paper has strong anti-motion deformation ability,which can solve the problems of small data volume,low frequency and large frame deformation of the equipment in this paper.Finally,through building indoor scene data set to test the algorithm in this paper,it is proved that the algorithm in this paper can be stable and robust for multi-scale and multi-functional indoor scenes The average relative translation error of the closed-loop trajectory is 1.095%,and the average relative rotation error is 0.0862 ° / m,which shows that the algorithm has strong anti-drift ability.The above experiments prove the effectiveness,stability and robustness of the algorithm,and show the feasibility of low-cost acquisition hardware for indoor 3D-slam.