From Quasi-Solutions to Solution: An Evolutionary Algorithm to Solve CSP

This paper describes an Evolutionary Algorithm that ree pairs to solve Constraint Satisfaction Problems. Knowledge about propp erties of the constraints network can permit to deene a tness function which is used to improve the stochastic search. A selection mechanism which exploits this tness function has been deened. The algorithm has been tested by running experiments on randomly generated 3-colouring graphs, with diierent constraints networks. We have also designed a specialized operator permutationn, which permits to improve the perforr mance of the classic crossover operator, reducing the generations number and a faster convergence to a global optimum, when the population is staying in a local optimum. The results suggest that the technique may be successfully applied to other CSP.