Research On LIDAR-based Environment Perception System For Unmanned Surface Vehicles

With the rapid development of information,communication,sensors and artificial intelligence technology,unmanned surface vehicles(USVs)are gradually playing an important role in marine environment monitoring,inland waterway measurement,automatic maritime patrol and unmanned shipping.However,only semi-autonomous USVs have normally been used,and they are not in a position to automatically perform complex missions under a variety of navigation environment.In autonomous navigation,the perception of the surrounding navigation environment is the fundamental aspect of auto navigation and obstacle avoidance,especially at close range.To improve auto-navigation ability of USVs,a LIDAR-based perception system for USVs is proposed.The main contents are as follows:(1)A LIDAR-based perception system for USVs is presented.The system consists of LIDAR subsystem,ship state subsystem and auxiliary subsystem.The parameters,operating principles,data features and processing methods of corresponding sensors are analyzed.(2)An obstacle detection and classification method for environment perception is proposed for USVs.In the Wuhan section of the Yangtze River,geometrical features obtained by LIDAR of obstacles such as vessels,piers and navigation marks are analyzed.These obstacles are simulated in the vessel pool by pool edge,pool corner and ball.Obstacles are preliminarily classified by filtering,adaptive distance threshold clustering and removal of isolated point sets.By using support vector machine,the characteristics in reflection point sets,including ball,edge,corner and unknown obstacles in the vessel pool,are defined for training SVM classifier.And the classification model is used to classify the four types of obstacles.(3)The software system is developed for USVs environment perception.Sensor data processing,obstacle classification,environment mapping and assistance modules are presented with the main interface and LIDAR interface.On VS 2010,each module is implemented.Details of the adaptive threshold clustering algorithm,SVM classification algorithm and environment mapping are also described here.(4)The performance test of the LIDAR-based environment perception system for USVs in the experimental environment was conducted.In the open-air ship pool,the experiment platform is established with environment mapping coordinate system for navigation and control.The system is tested in static and dynamic environments,showing the SVM obstacle classification combined with the adaptive distance threshold clustering and environment mapping are effective.The research and realization on the LIDAR-based environment perception system for USVs,have provided a reference for the environmental perception for USVs sailing in inland waterway,and enhance the autonomous navigation capability of USVs.