Abstract

Genetic Algorithms (GAs) can be used to solve difficult problems in many disciplines. The performance of GA depends on computer power. Parallel Genetic Algorithms (PGAs) are parallel implementations of GAs can provide better performance and scalability and can be implemented on networks of heterogeneous computers. Efficiency of GAs depends on crossover and mutation rates and it is difficult to adjust those parameters manually. This paper used multi agent techniques which combine existing GA and PGAs with distributed environment. The paper shows the efficiency of the parallel computation of the travelling salesman problem using the genetic approach on a multicomputer cluster. The master/slaves paradigm is applied. Performance has been made on the basis of MPI-based parallel program implementation. This algorithm is tested by the Traveling Salesman Problem (TSP). Result shows that, using of the parallel techniques can reduce the communication between different nods, therefore speed up the traditional genetic algorithm search process.

Keywords

The article was added to IASJ on 2016-07-01

204 Total full text downloads since the date of addition