Autonomous Navigation Method Of Autonomous Vehicles In Off-road Environment

The autonomous navigation of autonomous vehicles in off-road environment needs to realize 3D environmental perception and path planning,which is different from the 2D navigation in the urban environment.This paper studies the autonomous navigation method of autonomous vehicles in off-road environment for the main two difficulties.The detailed research contents are as follows:Firstly,this paper analyzes the demand of the autonomous navigation of autonomous vehicles in off-road environment.Then,the navigation scheme of autonomous vehicles is designed based on the demand.And the main sensors will be calibrated and the transformation equation between the main frames will be given.Meanwhile,the 3D kinematic model for vehicle will be set up.Thus,this part will build the foundation for the research of the key technologies.Secondly,the traditional stereo camera based methods have poor performance of time cost to reconstruct rough terrain in off-road environment.Therefore,this paper proposes a rapid method to build terrain grid map using V-disparity to solve the problem.The main idea of this method is to get the effective reconstruction regions and remove the nonaccessible regions from disparity image based on V-disparity.Then the effective reconstruction regions will be transformed to grid map through the stereo model.Thus the terrain elevation grid map can be rebuilt.Experimental results prove that the method could increase the efficiency and accuracy of reconstruction significantly since it only rebuilds effective reconstruction regions.Thirdly,the traditional LIDAR based methods can not reconstruct rough terrain in real time online.Therefore,this paper proposes a rapid method to reconstruct rough terrain based on Delaunay triangle to solve the problem.A large scale of environment point data can get from LIDAR.And a preprocess step and voxel based down sampling step will be done after get the point data.Then,the grid map will be get from the Delaunay triangle result based on Bowyer-Watson algorithm.Experimental results prove that the method can meet the requirement of navigation in a wide range of offroad environment for the automatic vehicles since it can reconstruct rough terrain in real time online.Finally,the traditional 2D path planning methods can not find the smaller undulating smooth path in off-road environment.Therefore,this paper proposes a 3D path planning method based on RRT to solve the problem.The first step is to use target biased and 3D sampling strategy to get the initial path.Then,the initial path will be simplified by pruning process based on maximum curvature.Last,the simplified path will be smooth by B-splines based method and get the executable non-collision smooth three-dimensional path with small undulations.Experimental results prove that the method owns good performance in time cost and can get more undulating path.