Algorithm Improvement Of Indoor Simultaneous Location And Mapping (SLAM) Based On LiDAR/INS Integration

The establishment of highly efficient,accurate and low-cost indoor mapping technology is becoming more and more necessary and urgent,due to the growing interest and market of indoor Location Based Services(LBSs).Multi-sensors(LiDAR/IMU/CAMERA)integrated indoor Simultaneous Location and Mapping(SLAM)technology becomes a promising solution,for no single navigation technology is robust enough to meet the requirements on its own.This paper builds a LiDAR/INS integrated based indoor mobile mapping platform.Based on the problems and shortcomings of the previous work,an improved LiDAR/INS integrated method is proposed,and the previous post-processed work was improved to a real-time solution in this paper.Finally,the methods are validated by a set of experiments.The main research work are summarized as follows:(1)The multi-resolution occupancy grid map and LiDAR scan matching algorithm are described in detail,and the time-consuming of two kinds of LiDAR scan matching algorithm—IMLE algorithm and Guass-Newton algorithm are compared.The experiment result shows that the mean time-consuming of the former is 64.2ms,the maximum is 105.5ms;while the latter is 2.0ms and 9.5ms respectively.Therefore,the Guass-Newton is chose for LiDAR scan matching is due to the low time-consuming,which not only can help save the total mapping time,but also provide convenience for the real-time performance improvement.(2)Although the positioning results can be achieved by using Guass-Newton scan matching in very short time,the shortcoming of this method is that it can easily get stuck in local minima if the initial searching value is not accurate.Aimed to the issue,this paper proposed an improved LiDAR/INS integrated algorithm to mitigate the problem.The positioning and mapping results of LiDAR/INS integrated method,only INS attitude based LiDAR scan matching and standalone LiDAR scan matching are compared and analyzed in experiments.The results shows that,compared with the high precision reference map,the mapping results of LiDAR/INS integrated method proposed in this paper was best.The RMS errors was about 0.0342m.Considering the errors brought by manual operation,the accuracy is reasonable;In the rich dynamic conditions,only INS attitude based LiDAR positioning results became inaccurate,and the errors of mapping results became larger and larger;However,even in the unabundant situation,the standalone LiDAR scan matching was also easy to get stuck in local optimum solution.Therefore,the LiDAR/INS integrated solution can help mitigate the problem,and the more abundant dynamic information,the better performance of INS.(3)The real-time response issue of multi-sensors is a big challenge for a real-time SLAM system,due to the different sampling frequencies and processing time of different algorithms.In this paper,an online Extended Kalman Filter(EKF)integrated algorithm of LiDAR scan matching and IMU mechanization for indoor mobile navigation system is introduced.Since LiDAR scan matching is considerably more time consuming than the IMU mechanism,the real-time synchronous issue is solved via a one-step-error-state-transition method in EKF.When the mobile mapping system is wandering,mapping results can be shown on the computer in the meantime.Compared to the traditional sequential post-processed EKF algorithm,the proposed method can significantly mitigate the time-delay of navigation outputs under the premise of guaranteeing the positioning accuracy,which can be used as an online navigation solution for indoor mobile mappingIn summary,this paper focuses on improving the LiDAR/INS integrated navigation algorithm to help Guass-Newton approach avoid get stuck in local minima;And improving the traditional post-processed method to a real-time solution.In addition,this research is also a push to use the MEMS inertial sensors in the indoor precision navigation and mapping.