Collision avoidance of moving obstacles for ship with genetic algorithm

Abstract

By use of a path planning method-genetic algorithms (GA), collisions with navigation obstacles and other moving ships can be avoided according to navigation traffic rule (the ship must pass on the right). Navigation obstacles and moving ships are identified by automatic radar plotting aids (ARPA), and then the future danger which is generated by the obstacles is predicted using a stochastic predictor. The stochastic predictor is designed such that the probability density map of the obstacle existence is predicted based on a Markov process model in enough future time. Evaluation of collision danger and efficiency of a path is taken in a cost function of GA. A safe and optimal path is given by this method. In heavy navigation traffic areas, correct selection of ship's speed, when accompanied by necessary path changes, is a crucial element in the safe navigation too. So the speed of own ship is introduced to the cost function of GA, and is modified along environment of navigation and the value of the cost function. The control method is confirmed by an experimental system that is the "Shioji Maru", which is the training ship of Tokyo University of Mercantile Marine. The experimental results are presented.