The uncertainty problem is one of problems that decelerate the establishment of the recovery network. Specifically, the returned products have less information on their: quantity, quality and location. This studied problem concerns the most crucial step during the implementation of a reverse supply chain: the collection process. To raise proportionally the uncertainty problem in large collection network and to give the decision about the location of collection centers and the allocation of the various collection points to these collection centers, a genetic algorithm tool of optimization is proposed. We propose also a mixed integer linear programming (MILP) and two new heuristics. We have compared between the differents techniques and the obtained results validate the efficiency of the proposed genetic algorithm .