Research On Indoor Autonomous Navigation Algorithm Of AGV Based On Graph Optimization Laser SLAM

The transportation of traditional materials is mainly composed of forklifts,trailers and other equipment.It requires a large number of personnel to participate in the work,but the labor cost is high and the efficiency is low,which is difficult to meet the future development needs of the enterprise.As a wheeled robot,AGV has been widely used as an important transportation equipment because of its wide working range and convenient deployment.In order to improve the flexibility of AGV operation and reduce the difficulty of on-site construction of the workshop,it is of great significance to study an autonomous navigation AGV to improve the automation of material transportation.This paper analyzes and summarizes the research status of SLAM-based autonomous navigation,studies the mechanical structure,hardware and software,and graph optimization algorithms of the AGV system,and builds the software and hardware system of AGV.The main research contents are as follows:The line segment features in the indoor structured environment are analyzed.For the PDBS algorithm,the corner points cannot be extracted during the region segmentation,and the line segment extraction failure caused by the inability to identify the breakpoints during the IEPF algorithm extraction process.A hybrid feature extraction of line segments combined with PDBS and IEPF methods is proposed.The algorithm uses PDBS,IEPF and hybrid algorithm to perform line segment feature extraction and actual environment comparison respectively.The improved algorithm extracts the number of corner points to 59 and the number of breakpoints is 4,which matches the environment.The extraction algorithm is closer to the actual environment,verifying the effectiveness of the hybrid algorithm.Based on graph theory and nonlinear least squares algorithm,the principle and basic framework of graph optimization algorithm are studied.Aiming at the problem of large computational complexity of ICP algorithm in pose solving process,a heading angle of gyroscope output is proposed instead of heading in ICP calculation process.The angle method,by performing accuracy and calculation time tests on the ICP algorithm and the improved ICP algorithm in six different poses in the environment,the calculation time used to improve the ICP algorithm is reduced from the average value before the improvement to 0.109 seconds to 0.075 seconds.The average value of the absolute value of the pose error decreased from(6.83,32.17,0.89°)to(3.50,12.83,0.46°).which verifies that the improved ICP algorithm is superior to the ICP algorithm in terms of speed and accuracy.Based on the graph optimization framework and ROS distributed characteristics,experiments were carried out in the "8m × 4m" environment and an 80m × 40m corridor environment in the laboratory.By comparing the pre-optimized and optimized maps,the same environment was used multiple times.The coincidence of the map obtained by laser scanning is improved.and the effect of graph optimization is verified.By comparing the line segments in the two environments and the length of the line in the actual environment,the average error of the length of the map segment length in the"8m × 4m" laboratory indoor environment is compared with the "80m × 40m" corridor.The environment is 2.95%small.The experiment shows that the number of features such as the inner segment of the environment is large and the error accumulation of the small odometer in the environmental range is small,which is beneficial to the positioning of the robot and the improvement of the map precision.