D3 G2A: A Dynamic Distributed Double Guided Genetic Algorithm for Optimization and Constraint Satisfaction

S. Bouamama and K. Ghedira (Tunisia)

Keywords

Multiagent systems, optimization, genetic algorithms, guidance, constraint satisfaction problem

Abstract

Inspired by the dynamic distributed guided genetic algorithm for Max-CSPs (D3 G2 A), D3 G2 A is a multi agent approach which addresses Constraint Satisfaction and optimization Problems (CSOPs).. In addition to GA efficiency providing good solution quality for CSOPs and to the benefits from multi-agent principles reducing GA temporal complexity, this approach possesses a new parameter called guidance operator and its agents have to determine their own GAs parameters. These latter allow not only diversification but also escaping from local optima. The new algorithm is, in fact, dynamic. It will let the species agent able to count its cross-over probabilities and its mutation probabilities. This operation is based on the species typology. Indeed, the sub-population of a species agent can be considered as strong or weak depending on its augmented fitness value.

Important Links:



Go Back