Research On Navigation Risk Analysis Methods For Autonomous Ships Under Incomplete Decision Information

When navigating in traffic congested areas and affected by complex conditions,traditional ships are more liable to have such accidents as collisions and fires causing casualties,serious property losses and environmental pollution.To enhance shipping safety and environment protection,and to reduce costs,intelligence levels of ships are growing continuously,and the scenario where autonomous ships and conventional ships sail in the same water areas will become a reality in the near future,which will pose new challenges for maritime administrations.At present,many problems arise when conducting risk analysis of autonomous ships,including shortage of historical accident data and that research methods cannot compensate for experts’ uncertainty and hesitancy in their evaluation making.Therefore,in this thesis,intuitionistic fuzzy sets and D-S evidence theory are utilized in research on navigation risk analysis of autonomous ships in such context of incomplete decision information.The details of the study are as follows:(1)In terms of hazard identification,based on the historical accident reports of traditional ships navigating in intensive waters,firstly,seven risk factors are summarized from the perspectives of human,ship,environment and management,and then they are mapped to the Class 3 autonomous ships;thereafter,in combination with the International Maritime Organization(IMO)autonomous ship risk analysis proposal and the relevant norms of international classification societies,18 risk factors of autonomous ships navigating in intensive waters are identified.Further,experts in R&D,design and operation of autonomous ships are invited to evaluate these risks identified.(2)With respect to questionnaire design,underpinned by expert evaluation,the incomplete decision information is converted into complete decision information.To reduce the subjectivity of expert evaluation,this thesis calculates the probability and consequence severity values of seven risk factors of traditional ships,which is referred to by experts in risk factors evaluation.Then entropy weight method is used to measure the uncertainty of the expert evaluation.(3)With regard to risk factors ranking,this thesis,based on the D-S traditional theory,and through analyzing the shortcomings of the method,introduces the intuitionistic fuzzy entropy and Jousselme distance to improve the method.The navigation risk factors of autonomous ships are ranked taking into account the treatment of uncertain information and conflicting evidence,using the likelihood of occurrence and severity of consequences of risk factors as evaluation indexes,and integrating experts’ evaluation information.Additionally,the feasibility and effectiveness of the proposed method are verified by comparing it with the traditional D-S theory and Sun Quan’s improvement method.The research findings indicate that such factors as fire prevention measures and equipment failure,power system failure,ship operating status perception and environment perception,unfavorable navigation environment,steering control failure,remote control system failure and collision avoidance decision failure are of higher risks and should be prioritized in the future accident tree and event tree analysis.This thesis uses a method combining objective historical data of traditional ships with the subjective data of expert evaluation,which provides a new insight into the research of autonomous ship risk analysis under the context of incomplete decision information and effectively solves the problems of evaluation information uncertainty and conflicting evidences.The ranking results of this thesis are basically consistent with the conclusions of related literature,which also verifies the reliability of this method.The method proposed in this thesis is expected to provide references for analyzing autonomous ships’ risks in other aspects,including cabin fires,cyber security,to name but a few.