Decision-Making Model for Multi-Ship Collision Avoidance Based on Adaptive Genetic Algorithm
Jiang-ling Hao, Li-ning Zhao, Jing-feng Hu, Xiaobo Yang · 2007
Decision-making for multi-snip collision avoidance usually concerns people's experience and judgment, so it is difficult to make a pure quantificational model. An adaptive genetic algorithm model is proposed in this paper. It introduces a kind of encoding method combining binary numbers and decimal numbers. In the model the solutions space consists of own ship's course and speed. For improving the performance speed and guaranteeing the precision, the integer part of course is represented with 9 binary numbers; the decimal part is represented with 1 decimal number. The integer part of speed is represented with 7 binary numbers; the decimal part is represented with 1 decimal number. Finally, a genome string, which combines course string and speed string, contains 18 numbers representing a solution. Considering sailing regulations, based on a collision-risk function which is formed with DCPA and TCPA, a fitness function is defined to decide whether an individual is selected or not. In order to ensure converging at an optimal solution, the model operates individuals with adaptive method. Both adaptive probability formulae of crossover and mutation are presented so that they can adjust themselves to individual's fitness. By these means, the model becomes more robust. The simulation results show that the model is able to achieve a correct action for collision avoidance.