Research On Automatic Decision-making System Of Vessel Collision Avoidance

The research on automatic decision-making system of vessel collision avoidance (ACAS) is an important topic in 21 century. Therefore, based on the summary and analysis on researches done before and on collision avoidance behaviors inquiry, using qualitative and quantitative analysis methods and at the view of subjective requisition united with objective requisition, the main results of this dissertation summarized as follows:Three encounter situations are divided quantitatively and the situations of collision avoidance action are also divided. Based on collision avoidance behavior inquiry and qualitative and quantitative analysis, three encounter situations are divided quantitatively by using fuzzy theory. The result absorbs in the mariners' knowledge for collision avoidance regulations and can improve the knowledge-base of AC AS.The concepts of space collision risk (SCR) and time collision risk (TCR) are fist put forward, the collision risk model of multiple factors and self-adjust is established. According to the connotation of DCPA and TCPA concepts, the concepts of SCR and TCP are proposed and the models are established separately. On the basis of stimulation-action principle, the compound operator of SCR and TCR is determined. The BP NN tool is used to calculate collision risk differently with others, it takes the row data obtained directly from ARPA as BP NN's inputs, the rapid calculation PB NN is determined and the results of BP NN simulation are very good.The last-helm model of multiple factors and self-adjusted is proposed. Based on the divisions of collision avoidance action situation, the last-helm model of multiple factors and self-adjusted is established in considering the factors of own-ship's speed, the speeds ratio of both encounter ships, encounter situation, own ship's turning property, the requirement of last-helm action, initial DCPA, the loss of distance during turning, the length of own ship. The characteristic of the model not only is multiple factors, but also includes the possibility of turning direction of the other ship. It adds a mechanism in ACAS to reduce or avoid non-coordination collision avoidance action.The entropy theory is used to analyze and evaluate the uncertainty ofcollision avoidance behaviors. It supplies a theoretical basis for the opportunity of collision avoidance decision-making. According to the investigation on mariner's behavior in collision avoidance and the relationship between uncertainty of collision avoidance behaviors and collision accidents, the entropy theory is used to evaluate the uncertainty in near head-on situations and crossing situations which are stand-on vessel crossing astern of given-way vessel initially.An optimization opportunity model is proposed for given-way vessel to take collision avoidance action. Based on the opportunity model of fuzzy satisfactory degree of stand-on vessel, the opportunity model of random program, the opportunity model of avoiding non-coordination collision avoidance action in considering the stand-on vessel may taking action, the last-helm model, the static state optimization model is proposed. Because the static state model does not reflect the changes of speeds of both encounter ships, therefore, the dynamic state model is put forward. The model established reflects not only the subjective requirement of stand-on vessel, but also the reality of given-vessel taking action in designated situation and unified the regulation's requirement of "made in amply time" with the mariner's knowledge.An optimization model of course alteration degree is proposed. The opportunity for given-way vessel to take action is a very important element for both ships to avoid non-coordination action of collision avoidance, but the degree of course alteration is the other too. It reflects the regulation's requirements of "the action shall be large enough to be readily apparent to another vessel observing visually or by radar". Therefore, the optimized model proposed by determining fuzzy functions of safety pas...