波浪中近距离船舶智能避碰辅助决策模型研究

In this project, the research on intelligent collision avoidance decision-making support model for ships in waves and at close quarters is carried out. Firstly, based on traditional geometric model of collision avoidance, the safe passing distance model of ships at close quarters is developed by considering the influence of ships’ main particulars on collision avoidance action. Secondly, on the basis of Manoeuvring Modelling Group (MMG) model, the unified theory and two-time scale models are applied to calculate the second-order wave force and moments when ship is maneuvering in waves, then the mathematical model of the ship maneuvering in waves is proposed. Thirdly, by utilizing adaptive control technique, such as neural network control and backstepping design, a robust adaptive control scheme is presented for ship maneuvering system in waves. Fourth, the related information is used to develop the dynamic calculation model of collision avoidance parameter, then the collision avoidance dynamic simulation model for ships in waves and at close quarters is achieved based on ship maneuverability in waves. At last, according to aforementioned research work, the intelligent collision avoidance decision-making support model for ships in waves and at close quarters is proposed by combining the International Regulations for Preventing Collisions at Sea and advanced Artificial Intelligence theory, such as Deep Reinforcement Learning and Broad Learning System. Moreover, simulation examples are employed to illustrate the effectiveness of the proposed decision-making support model. The decision-making support model is able to provide an effective and practical collision avoidance plan for ship’s officer, as well as lay the theoretic and technique foundation for autonomous collision avoidance and navigation of intelligent ship in the future.

本项目主要开展波浪中近距离船舶智能避碰辅助决策模型研究。首先在传统避碰几何模型基础上，进一步考虑船舶大小对近距离船舶避碰行动效果的影响，研究建立近距离船舶避碰安全通过距离模型；然后以MMG模型为基础，利用统一理论和双时间尺度数值模拟方法计算样本船舶在波浪中操纵时受到的二阶波浪力及力矩，给出波浪中船舶操纵运动数学模型；接着采用神经网络自适应控制、后推控制等先进控制理论和技术，设计在各种海况条件下都适用的船舶操纵运动控制算法；随后根据避碰必要信息构建船舶避碰要素动态计算模型，并结合波浪中船舶操纵性研究建立波浪中近距离船舶避碰动态仿真模型；最后基于上述研究，将国际海上避碰规则与深度强化学习、宽度学习系统等人工智能理论结合，研究建立波浪中近距离船舶智能避碰辅助决策模型，进行仿真分析研究，并验证其有效性。为当前船舶驾驶员提供实用可行的避碰策略，为未来智能船舶实现自主避碰及自主航行奠定理论与技术基础。

本项目主要开展波浪中近距离船舶智能避碰辅助决策模型研究。首先，基于重特大船舶碰撞事故资料，建立船舶碰撞事故深层致因分析模型，利用空间聚类分析方法、统计力学方法等建立船舶碰撞危险度预测和危险空间分布模型，并在传统避碰几何模型基础上，进一步考虑船舶大小对近距离船舶避碰行动效果的影响。然后，基于瞬态自由面格林函数提出一种新的规则波中船舶水动力三维时域面元法求解法，利用多系数保角变换法对船舶横剖面水动力系数进行求解，并采用STF法对全船水动力系数与波浪激励力进行积分，预测船舶在波浪中操纵运动数值及轨迹。接着，结合后推控制技术、隐函数定理、中值定理和动态面控制技术等方法，设计了一类考虑了风、浪等外界干扰的船舶操纵运动不确定性非仿射纯反馈非线性系统直接自适应神经网络控制算法，能够同时消除循环问题和复杂性爆炸问题，同时控制器结构十分简洁，控制效果优良，且便于实践应用。随后，根据避碰必要信息构建船舶避碰要素动态计算模型，并结合波浪中船舶操纵性提出一种充分考虑船舶运动特性的近距离会遇时船舶避碰动态辅助模型，能准确描述避碰过程中船舶操纵运动轨迹。最后，基于上述研究，将国际海上避碰规则与遗传算法、人工势场法等方法结合，研究建立多船实时协同避碰决策辅助算法，进行仿真分析研究，并验证其有效性。为当前船舶驾驶员提供实用可行的避碰策略，为未来智能船舶实现自主避碰及自主航行奠定理论与技术基础。