Collision avoidance control of ship with genetic algorithm

Concerns collision avoidance in ship navigation. Ships are usually equipped with ARPA systems, which process radar information and announce when other ships are near. An operator is still needed to watch the situation, make judgements and control the ship. The criteria of judgement are very complex, leading to misjudgements and mistakes. For a precise judgement and operation, theoretical criteria are not useful and a plenty of experience is required, but the captain or officer who has this ability are decreasing and necessity of automatic collision avoidance system is increasing. The authors adopt genetic algorithm (GA) for this. It has wide adaptability with learning function. The collision avoidance control system is constructed with following functions: 1) Get the information of other ships from the ARPA system; 2) Decide the navigating condition with GA 3) Control the ship direction with autopilot and ship speed with engine control. To confirm the performance of this system, the authors prepared three steps of experiment: 1) On board test using the ship Shioji-maru with virtual objects; 2) Real time simulation test with training simulator of Shioji-maru; 3) On board test using Shioji-maru with real objects. Basic functioning has been confirmed and various improvements for practical use are progressing.