In this paper some mixed techniques are outlined in order to combine the advantages of two very different methods for the resolution of combinatorial optimization problems (Genetic Algorithms and Branch and Bound Techniques), simultaneously avoiding their drawbacks. Due to the disparity between the basic techniques it is not suitable that they work at the same level so two models have been developed in which each technique assumes the role of being a tool of the other one. Parallelism is an important issue in those techniques.