Estimating ship collision avoidance probabilities

Mathematical approaches and tools used to predict collision frequencies of ships on a passage typically divide prediction into two separate likelihood (probability) terms. The first term represents the frequency of ship encounters with obstacles, either other ships or fixed structures, where the encounters lead to collision if no evasive action is taken. The second term represents the general ability of ships to take correct evasive actions to avoid the imminent collision. In multi-vessel encounter situations, in particular, the evasive actions may be complex interdependent actions by the involved ships. The paper introduces an approach for collision risk assessment based on Event Tree - modelling. The Event Tree - model depicts situation awareness, and subsequent actions of any two vessels in a multi-vessel encounter situation allowing assessment and comparison of risk control options. The estimation of the conditional probabilities of the Event Tree - model is specific to the encounter type which is defined according to the encounter geometries of the two vessels interacting to avoid collision. The probabilities can be estimated partly on existing track records of incidents and collisions, partly on dynamic vessel management simulation models.