船舶驾驶智能决策需处理环境中的不确定性信息，并满足规则与经济性的多重约束。模糊隶属度函数可量化这种不确定性并识别碰撞风险，结合PSO-GA算法可在规则约束下高效寻得安全与经济的最优避碰决策平衡解。故本研究基于模糊隶属度函数的船舶驾驶智能避碰决策方法。首先，通过分析船舶会遇态势，明确船舶避让责任；然后，构建融合舷角、船速比、DCPA与TCPA的模糊隶属度函数，以精准评估碰撞危险度（CRI）。当CRI超出阈值时，启动PSO-GA混合优化算法，以转向幅度与航行时间为决策变量，以安全性与经济性目标为适应度函数，搜索避让船舶的最优避碰航线，实现智能避碰决策。实验结果表明：应用该方法得到的避碰决策（右转30.5°，航行0.16 h）可将船舶的CRI降至0.03，平均最优适应度值可低至0.2458，且仅需10代即可收敛，所生成的决策安全、经济且符合规则，为船舶驾驶中智能避碰提供了有效解决方案。
Intelligent decision-making for ship navigation requires handling uncertain information in the environment and meeting multiple constraints of rules and economy. The fuzzy membership function can quantify this uncertainty and identify collision risks. Combined with the PSO-GA algorithm, it can efficiently find the optimal collision avoidance decision balance solution that balances safety and economy under rule constraints. Therefore, this study proposes an intelligent collision avoidance decision-making method for ship navigation based on fuzzy membership function. Firstly, by analyzing the situation of ship encounters, clarify the responsibility for ship avoidance; Then, a fuzzy membership function is constructed that integrates ship angle, ship speed ratio, DCPA, and TCPA to accurately evaluate collision risk (CRI). When the CRI exceeds the threshold, the PSO-GA hybrid optimization algorithm is activated, with steering amplitude and navigation time as decision variables, and safety and economic goals as fitness functions, to search for the optimal collision avoidance route for avoiding ships and achieve intelligent collision avoidance decision-making. The experimental results show that the collision avoidance decision obtained by applying this method (turning 30.5° right, sailing for 0.16 hours) can reduce the CRI of the ship to 0.03, and the average optimal fitness value can be as low as 0.2458. It only takes 10 generations to converge, and the generated decision is safe, economical, and compliant with rules, providing an effective solution for intelligent collision avoidance in ship navigation.
2026,48(4): 190-195 收稿日期：2025-12-3
DOI：10.3404/j.issn.1672-7649.2026.04.029
分类号：U675
作者简介：李志特(1979-),男,硕士,副教授,研究方向为航海技术、交通运输工程及航运教育
参考文献：
[1] 陶毅涵, 杜佳璐. 拥挤水域中的无人船智能避碰决策与航迹跟踪控制[J]. 控制与决策, 2025, 40(1): 214-222
TAO Y H, DU J L. Intelligent collision avoidance decision-making and trajectory tracking control for USVs in congested waters[J]. Control and Decision, 2025, 40(1): 214-222
[2] 隋丽蓉, 高曙, 何伟. 基于多智能体深度强化学习的船舶协同避碰策略[J]. 控制与决策, 2023, 38(5): 1395-1402
SUI L R, GAO S, HE W. Ship cooperative collision avoidance strategy based on multi-agent deep reinforcement learning[J]. Control and Decision, 2023, 38(5): 1395-1402
[3] 关巍, 罗文哲, 崔哲闻. 基于深度强化学习的无人驾驶船舶避碰行为决策方法[J]. 大连海事大学学报, 2024, 50(1): 11-19
GUAN W, LUO We Z, CUI Z W. Decision making method for collision avoidance behavior of unmanned ships based on deep reinforcement learning[J]. Journal of Dalian Maritime University, 2024, 50(1): 11-19
[4] 张立华, 周寅飞, 贾帅东, 等. 一种有效顾及复杂海域避碰的路径规划方法[J]. 哈尔滨工程大学学报, 2023, 44(1): 56-64
ZHANG L H, ZHOU Y F, JIA S D, et al. A path planning method for collision avoidance of ships in complex sea areas[J]. Journal of Harbin Engineering University, 2023, 44(1): 56-64
[5] 徐言民, 律建辉, 刘佳仑, 等. 基于CSSOA的多船智能避碰决策研究[J]. 中国舰船研究, 2023, 18(6): 88-96
XU Y M, LYU J H, LIU J L, et al. Multi-vessel intelligent collision avoidance decision-making based on CSSOA[J]. Chinese Journal of Ship Research, 2023, 18(6): 88-96
[6] 张可, 黄立文, 贺益雄, 等. 基于航迹推演的船舶动态智能避碰方法[J]. 中国航海, 2023, 46(4): 20-29
ZHANG K, HUANG L W, HE Y X, et al. Dynamic intelligent collision avoidance method based on trajectory prediction[J]. Navigation of China, 2023, 46(4): 20-29
[7] 米佳林, 郑中义, 刘子豪. 基于复杂网络的船舶碰撞危险量化模型[J]. 中国航海, 2025, 48(1): 26-33,68
MI J L, ZHENG Z Y, LIU Z H. Quantitative modeling of ship collision hazards based on complex networks[J]. Navigation of China, 2025, 48(1): 26-33,68
[8] 陈兆彤, 陈江平, 潘励. 基于模糊隶属度的船舶中断航迹关联识别方法[J]. 应用科学学报, 2023, 41(2): 296-310
CHEN Z T, CHEN J P, PAN L. Identification method for vessel interrupt track correlating based on fuzzy membership degree[J]. Journal of Applied Sciences, 2023, 41(2): 296-310
