Hanafi Alamdari - Electrical Engineering Faculty University Of Mazandran
Miar Naimi - Electrical Engineering Faculty University Of Mazandran

In this paper a new genetic algorithm is designed to solve a class of maze problems, more efficiently than the state of the arts. Here the solution is a set of paths directing every agent from any position in the maze towards a single goal point. The novel points of the proposed algorithm are the innovative crossover and mutation operators that improved the results. To evaluate the algorithm and compare it with other algorithms, different experiments were done. Here each algorithm for any problem was run five times and the average of the results was considered as the evaluation measure. All experiments showed that the proposed algorithm is the most efficient one.

Maze, genetic algorithm, multi agent problems