Planning and keeping the safe course to avoid collision at sea with genetic algorithms

According to navigating traffic rules (e.g. it must pass the other ship from the right), navigating obstacles and other moving ships can be avoided using genetic algorithms. The navigating obstacles are identified by automatic radar plotting aids (ARPA) system. Then the future danger generated by the obstacles can be predicted using a stochastic predictor. Evaluation of collision danger and efficiency of a path is taken in a cost function. To avoid the navigating obstacles, the original course has to be changed with the change of head direction. So keeping the course to guarantee her coving the given route to the goal place on time is also most important. For this purpose, the head direction operator is introduced to the cost function and her course error is considered in ending gene of every chromosomes. This control system is examined by "Shioji Maru", which is the training ship of our university. The experiment results are presented.