| With the rapid development of global economy and trade,the scale of the maritime freight industry is also expanding,and the density of ships at sea has increased sharply,resulting in frequent ship collisions.Most of the ship collision accidents are caused by the inattention of the ship driver or the lack of driving skills.The ship collision avoidance assistant decision-making system can help the driver to complete the collision avoidance assistant driving.The current related research has problems such as incomplete coverage of collision avoidance scenarios and unclear definition of collision avoidance behavior.Therefore,on the basis of comprehensively constructing the knowledge base of ship collision avoidance,this paper designs a ship intelligent collision avoidance algorithm based on reinforcement learning,which makes the decision information provided by the ship collision avoidance auxiliary decision-making system more reference and practical.Firstly,the basic principles of the mathematical modeling of ship collision avoidance are explained,including the ship reference coordinate system,the mathematical model of ship motion,the mathematical model of ship maneuvering,and the mathematical model of ship collision avoidance,modeling.The space collision risk and time collision risk are quantitatively analyzed under the ship encounter situation,and the judgment method of the ship collision risk is given.Secondly,according to the different sources of ship collision avoidance knowledge,the collected knowledge in the field of ship collision avoidance is summarized and sorted,and divided into knowledge of maritime collision avoidance rules,knowledge of ship maneuverability,knowledge of collision avoidance stage division,meeting situation division knowledge,There are several categories of emergency ship operation knowledge and navigation experience knowledge.Then,knowledge engineering methods such as frame knowledge representation and process knowledge representation are used to express different types of ship collision avoidance knowledge,and on this basis,a knowledge base for ship collision avoidance is constructed.Then,after establishing the mathematical model of collision avoidance,a quantitative analysis of the key elements of ship collision avoidance is carried out,and an intelligent ship collision avoidance algorithm based on reinforcement learning is proposed.Based on the near-end strategy optimization algorithm,according to the actual situation of ship collision avoidance at sea,the ship collision avoidance environment state set,the collision avoidance behavior action set and the collision avoidance excitation function are designed respectively.And after building the knowledge base of ship collision avoidance,the ship collision avoidance rules are embedded in the intelligent collision avoidance algorithm,so that the ship agent can make reasonable avoidance behaviors according to the constraints of the sea collision avoidance rules.The ship intelligent collision avoidance algorithm is simulated and verified in three typical encounter scenarios of ship encounter,cross encounter and overtaking.The results show that the ship simulation model trained by the ship intelligent collision avoidance algorithm can make correct avoidance behavior and avoid The nearest encounter distance index of the latter two ships is greater than the requirement of the safe encounter distance,and at the same time,it can ensure that the ship turns correctly after the collision avoidance behavior ends and returns to the preset route.Finally,in the Windows 10 environment,the development tool of Qt 5.9.0 is used to complete the construction of the auxiliary decision-making system for ship collision avoidance.function,collision risk calculation function and ship encounter situation judgment function. |
