Cognitive Techniques For Concave Barriers In Unstructured Environments

It is one of the key technologies in the field of mobile robotics research to implement obstacle detection and barrier awareness in the environment whose information is non-fixed,cannot be known,unstable unstructured,and the structure and size changes irregularly,and it is also the necessary basis for robots to complete other advanced tasks.Nowadays,a lot of research has been done about the obstacles detection at home and abroad,but most of them are aimed at convex obstacles,and there are few studies on concave obstacles.In this paper,building a mobile robot platform which is based on the main visual sensor a three-dimensional lidar,puts forward a laser sensor multi-information feature fusion recognition algorithm to complete the detection of hidden ditch or others and accomplishes the detection of concave obstacles under the unstructured environment.Firstly,construction of mobile robot system platform;compared to different types of laser radar,for this study completed the laser radar selection work.Secondly,based on the characteristics of point cloud data,a filtering method based on curvature classification is proposed for point cloud data obtained by multi-line lidar.The average Gaussian curvature value in the neighborhood of the sampling point is used to divide the region type of the point cloud data model,and then adaptive median filtering and adaptive two-sided filtering are used for different regional types which make it suitable for both static target environment and dynamic target process.Thirdly,aiming at the various obstacles in unstructured environment,the research is carried out.Firstly,the detection of the laser return strength value is measured for the water barrier.Secondly,the detection of the vegetation type obstacle is completed based on the spatial consistency.Finally,the detection of the concave barrier is achieved by abrupt changes in radial spacing between adjacent laser beams.Fourthly,based on the analysis of the accessibility of the mobile platform,a mathematic model of the scanning line is established.Aiming at the shortcomings of the traditional vertical installation of lidar and the sparsely populated data density,this designs a new multi-cell laser radar layout method to improve the point cloud density in the ROI,the mobile robot in front of the blind area control within 1.5 meters;Fifthly,this paper proposes a laser-sensing multi-information feature fusion algorithm based on Bayesian criterion to assign different weights to complete the detection problem of concave barrier in unstructured environment.Based on the detection of concave obstacle,experiments have been done about the mobile robot platform,and the experimental results show that the method can be used to identify the concave barrier in the range of 2 to 20 meters in the unstructured environment,and realize the cognitive task.