华中科技大学学报（自然科学版）2025，Vol.53Issue(4)：15-21,7.DOI:10.13245/j.hust.250115
面向智能航行测试的船舶会遇场景案例生成方法
Ship encounter scenario case generation method for intelligent navigation testing
摘要
Abstract
To evaluate the performance of autonomous ship collision avoidance and intelligent control algorithms under different risk scenarios,a data-driven generation method for ship encounter scenarios was proposed.A large number of encounter test scenarios were randomly generated in the quaternion domain of ships,and the Fuzzy-CRI(Fuzzy-Collision Risk Index)was employed to pre-classify the scenario data.A scenario data compression network model was constructed based on the BP(Back Propagation)neural network,and reverse training was conducted on the classified data to derive ship encounter scenarios under different risk levels.Given different risk levels and encounter situations,the model was capable of automatically generating test scenarios.The Fuzzy-CRI was employed again to verify the training results of the model.Results show that the accuracy of the self-generated scenario model exceeds 85%.Moreover,the model occupies minimal storage space and demonstrates high retrieval efficiency,which could provide test scenarios with varying risk levels for intelligent navigation planning and control algorithm.关键词
船舶测试验证/场景生成/船舶会遇态势/模糊理论/神经网络Key words
ship testing and verification/scenario generation/ship encounter situation/fuzzy theory/neural network分类
交通工程引用本文复制引用
刘佳仑,刘东豪,李诗杰,胡欣珏..面向智能航行测试的船舶会遇场景案例生成方法[J].华中科技大学学报（自然科学版）,2025,53(4):15-21,7.基金项目
国家重点研发计划资助项目(2022YFB4301402) （2022YFB4301402）
国家自然科学基金资助项目(52272425). （52272425）
