Real-time Positioning of Mobile Robots Based on Improved 3D Normal Distributions Transform Algorithm
To address the issues of low registration accuracy and long registration time in 3D normal distributions transform(3D-NDT)point cloud registration when initial poses are not accurately known,which fail to meet the real-time localization requirements of mobile robots,an improved 3D-NDT point cloud registration fusion algorithm was proposed.During the downsampling process of the raw point cloud,points from the source point cloud were used to replace the calculated centroids,thereby reducing computational complexity while preserving the feature information of the point cloud.By introducing a trust radius to dynamically adjust the iteration step size,the accuracy after downsampling and the speed of point cloud registration could be improved.Additionally,by integrating 3D laser point cloud data with 9-axis inertial measurement unit(IMU)data,problems such as excessive pose differences between two sets of point cloud data leading to non-convergence or falling into local minima were resolved.The improved 3D-NDT algorithm was subjected to simulation experiments using a self-built mobile robot platform in the laboratory to verify the reliability and accuracy of the real-time localization of the enhanced algorithm.The results show that compared with the traditional 3D-NDT algorithm,the improved 3D-NDT algorithm achieves a matching accuracy improvement of 106%outdoors and 108%indoors,with success rates increasing by 8.29%and 6.35%,respectively.Average matching times were reduced by 51.1%and 47.9%,respectively.This significant enhancement in registration accuracy and substantial reduction in single-registration time for mobile robot real-time localization indicate that the improved 3D-NDT algorithm can meet the demands of real-time positioning for mobile robots.
normal distribution stransformpoint cloud datapoint cloud registrationinertial measurement unit(IMU)data fusiondynamic trust radiusautonomous localizationmobile robots
万方数据
维普
