Research On Environment Perception And 3D Path Planning Of Ground Unmanned Platform In Unstructured Scenarios

As a special form of unmanned vehicle,unmanned ground platform has received extensive attention and rapid development since it came into being.With the progress of modern military technology,unmanned ground platform as a reliable form of equipment,more and more countries become the focus of research.Different from civilian unmanned platforms operating in structured environments mainly in cities or villages,military unmanned platforms mostly operate in unstructured environments such as off-roads,which are characterized by complex terrain and diverse obstacles,which bring challenges to the operation of unmanned ground platforms,especially in the autonomous navigation system.Since most of the existing technologies are based on the flat environment,it is difficult to simply transplant to the unstructured environment.Therefore,in view of the barriers of existing technologies,this paper focuses on the path planning technology under the unstructured scenario of unmanned ground platform,which mainly includes the following aspects:(1)Aiming at the sensing needs of ground unmanned platforms in unstructured environments,the sensor scheme for environmental sensing in this paper is determined.After the coordinate system is unified,the effective detection range delimitation and data filtering are carried out for the target point cloud characteristics in the unstructured scene to reduce the amount of data processing.This paper designes an improved ICP fine registration algorithm based on kd-tree nearest neighbor search to complete the fusion of point cloud images of different frames,and combines the SAC-IA coarse registration algorithm to complete the registration of the 3D point cloud in the simulation experiment.(2)Research on unstructured scene 3-D environment modeling scheme,to digital elevation model as the expression of topography,to occupy the grid map as a map representation method,combining with the characteristics of both,and improved respectively,fusion contains the elevation dimension information of class 3-D environment model,put forward the concept of "non steady state obstacles" to the environment may be an obstacle to the potential effects on the ground unmanned platform driving factors classification definition,design the analysis extraction method.(3)Research on the three-dimensional path planning method,baseing on the environment model containing non steady state obstacle.Evaluate the ground unmanned platform crosscountry performance,increase in the cost function of A * algorithm degree evaluation of traffic and terrain penalty function.By setting up the simulation experiment,the algorithm has been verified that it has good usability,and can plan out a "shortcut" for its passage for the ground unmanned platform.(4)The actual vehicle verification system is designed based on the experimental car platform.The experimental car platform mainly includes the drive system,the on-board industrial computer,and the lidar.Set up a simulated unstructured experimental scene,first complete the environment perception and modeling of the experimental scene through hardware such as the ground unmanned experimental car platform and on-board lidar,and use the threedimensional path planning method proposed in this paper in the upper computer based on the environment model.Offline planning of two car platform states.Finally,based on the planned path,the path coordinates and control signals are sent to the execution layer of the chassis to realize the tracking execution of the path.Experimental results show that the environment perception and modeling method and 3-D path planning algorithm proposed in this paper can plan a suitable 3-D path in unstructured scenes.