Research On Lidar SLAM Algorithm For Indoor AGVs Based On Graph Optimization

With the gradual intelligentization of the warehousing AGVs(Automatic Guided Vehicle)in the industrial field,auto-navigation has gradually become an indispensable ability for industrial robots.The SLAM(Simultaneous Localization and Mapping)technology can help AGVs to independently complete positioning and the establishment of environmental maps without prior knowledge.Therefore,the research on SLAM technology for AGVs has important theoretical and practical significance.Aiming at the problem of inefficient point cloud matching and inaccurate positioning of a single sensor,this thesis studies the lidar SLAM based on graph optimization.Combined with the laboratory's self-developed AGV platform,the AGV positioning and mapping algorithm is designed and implemented.The specific research contents are as follows:(1)According to AGV design requirements,a distributed design of the AGV system is carried out,and Solidworks is used to simulate the car body.Meanwhile,a system model of AGV and sensors is established and the update principle and process of the grid map is derived.(2)Then,this thesis conducts a theoretical analysis of the SLAM algorithm based on the graph optimization,and makes related improvements on this basis.In the front-end point cloud matching part,a CSM(Correlation Scan Match)method combining grid resolution and angular resolution is proposed to solve the problem of noisy points and inefficient search,and a screening strategy of the points to be matched is added based on distance threshold.In the back-end optimization part,in view of the limited accuracy of laser point cloud matching and positioning,this thesis poses a back-end calibration and positioning method based on multi-sensor fusion,which adopts the ARIMA(Autoregressive Integrated Moving Average model)model to fit the sensor error.Besides,an improved XGboost method is proposed to perform positioning fusion of multiple sensors,which adds the fusion positioning result to the node of graph optimization to complete the optimization process.(3)At last,the AGV platform independently built by the laboratory is applied to verify the algorithm.At the same time,the open source data set,simulation environment and real physical environment are used to conduct experimental analysis on the proposed SLAM algorithm,which proves the effectiveness of the method.