Research On Virtual Testing And Evaluation Of Autonomous Navigation Algorithms Of USV

In recent years,with the continuous development of intelligent technology,the research of unmanned surface boats has received more and more attention from scholars.Unmanned boat autonomous navigation technology is the focus of research on unmanned boat technology,which is the key technology to ensuring the safe and efficient navigation of unmanned boats,and must be tested and evaluated in large quantities before being put into practical application.The testing and evaluation technology of the autonomous navigation capability of unmanned boats in China is still in the initial stage and has a large research space.In response to this situation,this thesis conducts research on the testing and evaluation technology of autonomous navigation algorithms.The main research contents are as follows.First,the autonomous navigation capability of unmanned boats is analyzed,and a virtual testing method for the autonomous navigation algorithm is proposed.Through this method,the autonomous navigation algorithm can be simulated and tested in four aspects,such as perception recognition,path planning,collision avoidance decision,and navigation control.At the same time,an unmanned boat motion model is established.In terms of virtual test scenarios,various test element models are established from the actual navigation environment of the unmanned boat,and three kinds of obstacle ship models are established from the test perspective.Based on the established test element models and obstacle ship models,four unit capability test scenarios and three comprehensive capability test scenarios were established from multiple test perspectives,providing a multi-angle and multilevel test means for the autonomous navigation algorithm.In terms of 3D simulation,the real navigation environment of the unmanned boat is simulated and diverse environmental changes are added.The virtual dataset labeling method and the virtual dataset creation method with rich changes are proposed for the maritime target,which provides a virtual test means for the perception and recognition capability of unmanned boats.In the evaluation of autonomous navigation capability,evaluation indexes are selected for the test targets and the definition and calculation methods of the indexes are introduced,and the evaluation indexes are made consistent.An autonomous navigation capability evaluation model is established to provide a basis for the evaluation of the autonomous navigation algorithm of unmanned boats.Finally,based on the above research,a virtual test platform,test data recording terminal,and test data display terminal are built.Based on the data information of the unmanned boat "Blue Letter",the test platform simulation verification and collision avoidance navigation algorithm simulation tests were conducted and the test results were evaluated.The results proved the feasibility of the proposed test method and the effectiveness of the evaluation method.