An Approach of Ship Routing Plan Based on Support Vector Machine

This paper focuses on a new approach of applying a pattern classification technique to ship routing plan. A safe path between a start point and a destination provides information about the space region partition. In the case of 2D routing plan, the route classifies the space into two districts. This means a dual problem of first classifying the entire space into two districts and then picking out the border as a route. We propose a novel approach to solve this dual problem based on support vector machine (SVM). SVM produces a non-linear separating surface on the basis of the margin maximization principle. This feature is applied to the objective of common routing plan problems, that is, generating a non-collision and smooth route. The effectiveness of the proposed approach is demonstrated by using several routing plan results in 2D spaces.