| With the increasing demand of ocean development and military application,Unmanned surface vessel has been widely concerned with its unique advantages and is an important equipment for carrying out complex military and civil tasks.In order to complete various specific tasks in complex waters,Unmanned surface vessel are required to have the ability of perceiving the surrounding complex water environment,generating obstacle-avoiding routes and avoiding maneuvering targets according to the water environment.In view of the above requirements,this paper carried out researches on lidar based surface target detection,local path planning and collision avoidance methods.The main works include:(1)Aiming at the problem that the detection results of traditional Euclidean clustering algorithm are sensitive to the distance threshold when detecting water surface targets based on laser point cloud,and the improper selection of the distance threshold is easy to cause under-segmentation or over-segmentation of targets,an improved Euclidean clustering algorithm is proposed in this paper.When the range threshold is large,the algorithm can distinguish different kinds of laser points in the search process by setting different weights of laser points for the cluster target and the interference target,so that the Euclidean clustering has a good clustering effect within a certain range of the range threshold,reducing the difficulty of threshold selection.In addition,in order to improve the real-time performance of clustering detection,an attribute grid clustering algorithm is further proposed in this paper,which converts point cloud data into two-dimensional attribute grid,obtains grid attributes according to laser points in the grid,and uses the attribute grid as the basic clustering unit for clustering.Finally,the effectiveness of the proposed algorithm is verified by comparing the proposed algorithm with the common water area point cloud clustering algorithm.(2)Based on the results of lidar target detection obtained by the above algorithms,path planning in local static environment is studied in this paper.Firstly,the static target information obtained from the clustering is used to model the water environment.On this basis,aiming at the local minimum problem of the artificial potential field method and the chattering phenomenon near large obstacles,an enhanced artificial potential field local path planning algorithm is proposed.This algorithm improves the existing artificial potential field by introducing the "ring force" potential field of large obstacles.Then,the feasible path is reprogrammed to reduce the path length,and Beizer curve is used to smooth the path,so that the final path is suitable for the actual navigation of Unmanned surface vessel.Finally,the effectiveness of the proposed path planning algorithm is verified by the perception data of the actual water body static scene.(3)Aiming at the common collision avoidance problem of dynamic targets in actual waters,the effective dynamic targets are determined and tracked from the detection results of lidar based on target life cycle decision and Kalman filter,and their motion state is estimated.Secondly,a DWA-VO(Dynamic Window Approach-Velocity Obstacle)collision avoidance algorithm with uncertain velocity is proposed.The algorithm considers the uncertainty of target state acquired by lidar.On the basis of DWA algorithm,the original evaluation function is optimized by combining speed obstacle method and navigation rules,and a new evaluation function is established.Finally,according to the change of collision avoidance risk during navigation,the proposed algorithm is used to make collision avoidance decision,and the algorithm is verified based on the perception data in the actual dynamic water environment. |
