Research On Laser SLAM Of Differential Mobile Robot Based On Graph Optimization

At present,intelligent mobile robots play an important role in people's daily lives.SLAM(Simultaneous Location and Mapping)technology is the core key technology in the development of mobile robots.Because lidar can obtain accurate distance information of obstacles,has strong anti-jamming ability,can build accurate map and well meet the construction of two-dimensional indoor environment,so laser sensor is selected as external measurement sensor to study slam technology.SLAM algorithm based on graph optimization can be established because of its high loop detection ability and real-time performance More accurate map,is the main research direction of laser SLAM technology.This paper takes SLAM based on graph optimization as the research basis,and relies on the ALEbot R&D platform of Shandong University.And this paper studies and optimizes the key technologies involved in graph optimization slam,builds an experimental platform,and conducts algorithm experiments in the real environment.The results show that the effect of graph construction is good.The main research content of this article is divided into the following aspects:(1)According to ALEbot research and development platform,the lidar model is established according to the principle of laser sensor ranging,and the motion model of the mobile robot is established according to the principle of differential drive of the R&D platform.The gyroscope can obtain more accurate rotation measurement characteristics,and improve the track estimation algorithm.Experiments show that the improved track estimation can achieve higher positioning accuracy at short distances,but it will exist in the face of longer operation.Cumulative errors lead to large positioning errors.(2)Aiming at the problem of erroneous matching during the inter-frame matching by the iterative closest point algorithm,the improved algorithm of machine learning and its fusion is used.The average absolute error of the improved inter frame matching algorithm in the X direction is 2.1863mm,the average absolute error in the Y direction is 2.6954mm,and the absolute angle error is 0.0645°.In order to avoid large positioning error caused by wheel slip,a laser odometer is established based on this.According to the data obtained from frame matching,the structural dimension of the wheel odometer is calibrated by using the model-based method.Through multiple experiments,the left wheel diameter of the wheel odometer is 144.344mm,the right wheel diameter is 143.875mm,and the wheel spacing is 525.833mm.(3)According to the map requirements in user visualization and subsequent navigation,the grid map is selected as the final presentation form of the final experimental map building;in view of the dependence on the laser odometer in the process of sub map building,The design combines the laser sensor,wheel encoder and IMU with the extended Kalman filter algorithm;Aiming at the problem of more sensitive to parameters in the feature extraction process of Split-Merge(SM)algorithm,the slope based method is proposed It can effectively extract the line features in laser point cloud by using the SM algorithm.For the loop detection of graph optimization,a map to map detection method is proposed.The "strong"constraint is established by loop detection,and the goal of correcting the robot's pose is achieved by using graph optimization algorithm.Build experimental platform to conduct experiments in an indoor environment.By comparing the real environment and the size of the map based on the optimized SLAM algorithm based on the map,the root mean square error in the map building is 0.0613m,which can meet the accuracy requirements of the map for mobile robots to navigate indoors.