出版社：The Japan Society of Naval Architects and Ocean Engineers
摘要：Growing the demand for new generation of automation in ship navigation and for various assessments in harbours and waterways, so-called fast-time simulation technique becomes important as well as man-in-the-loop simulation such as ship handling simulator experiments.In the fast-time simulation technique, the mathematical model of the ship dynamics in various environmental conditions should be of course carefully investigated, but the modelling of navigators is another important factor which governs the inputs of simulation. In this paper the authors have developed an automatic navigation system called SAFES in combination of fuzzy theory and expert system, which is applicable to automation of ship navigation and also to the modelling of navigators in the fast-time simulation. The system can treat various ship operations in open sea and restricted waterways as well as under multi-ship encounters. Some results of the application of SAFES are shown, which were actually used for certain assessment and design of waterways.In the modelling of ship navigators, fuzzy control and reasoning are used for the basic controls such as course-keeping, course-changing and path-keeping, and for the reasoning of collision risk between other ships and boundary of navigational lane. Further in the path-keeping algorythm, fuzzy reasoning of the nearness to the course-changing point is included. It is not necessary to introduce fuzzy control even in course-keeping, but it is found that fuzzy autopilot can operate smoothly from course-keeping to course-changing and vice versa without large overshoot nor slow settling. The describable knowledge of ship navigators for ship operation and process control is represented in the rule base of the expert system.As conclusions of the paper, the authors can point out the following two points as well as the availability of the proposed system. As the first point, the expert system approach is essential for the description of complicated navigators' decision process, which cannot be easily implimented in the sequential programming style such as in FORTRAN. On the contrary, each function of navigators' orders and behaviour of helmsmen may be represented also by non-fuzzy description. However, the authors would like to stress as the second point that once we prepare a general-purpose fuzzy controller/reasoner which passes one fuzzy output from two fuzzy inputs, almost all functions can be easily handled by fuzzy control/reasoning, because most functions of human operation are originally given by only fuzzy descriptions.
