Study On Collision Avoidance Decision Making Methodology Under Multi-ship Encounter Situations

In recent years,with the comprehensive implementation of the national “Belt and Road” strategy,the "transportation power" strategy,the "ocean power" strategy and the recovery and growth of the shipping economy,maritime transportation will continue to play a significant role in international trade.However,with the development of the ship in the direction of "large-scale hull,intelligent operation,automatic decision-making,and unmanned driving",serious ship collision accidents still occur occassionaly,causing catastrophic loss of life,property and environmental pollution.The intelligent collision avoidance technology is one of the key technologies for ship intelligent navigation.Therefore,researches on intelligent collision avoidance decision-making methods for ships,especially in compliance with the rules of the COLREGs(Convention on the International Regulation for Preventing Collisions at Sea),and improve the level of intelligent anti-collision decision-making performance under multi-ship encounter situations are among the most critical topics in the maritime transportation.This thesis focuses on some key issues in collision avoidance decision procedure,including multi-ship collision avoidance decision-making method in open waters,quantitative evaluation of ship collision risk,the logical code of the key provisions of the COLREGs and the heuristic algorithm for ship intelligent collision avoidance decisions-making.Specifically,the main research work is as follows:(1)Quantification of ship collision risk based on artificial neural network(ANN).Based on a comprehensiv analysis of the collision avoidance mechanism and the collision avoidance process of the ship,a calculation model of ship collision risk based on generalized regression neural network(GRNN)is constructed.GRNN has better nonlinear mapping ability and high fault tolerance than traditional BP neural network.The distance to closest point of approach(DCPA),time to closest point of approach(TCPA),relative distance,relative bearing and ship speed ratio are taken into consideration.By applying the fuzzy comprehensive evaluation method to the GRNN model,the advantages of the two methods are integrated through the neural network.The continuous learning is performed to obtain the collision risk index.The training results show that the error between the predicted value and the expected value of the GRNN model is less than 1%,and the generalization ability of the model is strong.So as to effectively quantify the collision risk between the own ship and the target ship,and provide reference for seafarers to make effective situation awareness.(2)Formulation of the logical code of COLREGs.On the basis of the systematic analysis of the rules using machine learning algorithm and the inductive logic programming method(ILP),the key clauses concerning ship safety and collision avoidance in COLREGs are formulated.A framework of general rule system is established.The collision avoidance rules are summarized into a program language that can be understood by the machine.The collision avoidance rules realized by logical semantic transfering can provide decision-making basis for subsequent collision avoidance algorithm,The algorithm can help realize effective determination on the encounter situation and the responsibilities of the encounter ships.(3)Decision-making methodology under multi-ship encounter situations based on heuristic algorithm.Firstly,the classical heuristic algorithm,particle swarm optimization(PSO)and genetic algorithm(GA)are used.By introducing the mutation and crossover operation in GA,the PSO is improved and the hybrid optimization algorithm model of PSO and GA are constructed.Secondly,according to the COLREGs and the requirements of good seamanship,the objective function of collision avoidance decision-making based on turning range and sailing time is constructed considering the safety and economy of ship navigation.The PSO-GA algorithm can effectively improve the convergence speed,optimization precision and stability.Finally,when the collision risk between ships reaches or exceeds the threshold value,the activation of collision avoidance algorithm is activated to complete the optimization of collision avoidance strategy for ships,and the PSO-GA algorithm is activated to search the optimal steering amplitude and the sailing time from the solution space.The simulations are carried out to check the feasibility and effectiveness of the algorithm under multi-ship encounter situations.The simulation results indicate that on multiship encounter situations,the PSO-GA can find an effective route with smaller iterations than PSO and GA algorithms for the give-way ships when the ship obeys the COLREGs or some ships violate the COLREGs.The approach can not only provide with good reference for seafarers to make collision avoidance decisions but also can reduce the risk of shipping traffic and improve the safety of navigation.To sum up,the ship collision risk calculation method and collision avoidance decision-making methodology proposed in this thesis can effectively quantify the collision risk and determine a reasonable time for collision avoidance decision-making.It can also use the programmatic logic coding method to achieve the appropriate judgment of encounter situation as well as their responsibilities.The relevant research results indicate that the proposed maneuvering plans guarantee the safe encountering of ships in an economical way.It can also provide with good reference for seafarers to make collision avoidance decisions.