There are a number of studies on the access network design and the fiber-based backbone network design. However, the results of which cannot be directly applied to the fiber-based access network design because of little consideration for the unique characteristic of it. The fiber-based access network design though similar to copper-based one, requires additional consideration such as ONU location.
So this thesis focuses on the problem determining the ONU location and capacity under given demand, which is important in FTTx design. For this purpose existed models, algorithms and implementation techniques for Local Access Telecommunication Network (LATN) are referred. We regard this problem as the General Facility Location (GFL) on trees. GFL can be reformulated Tree Partitioning Problem (TPP) with an exponential number of columns. We can solve the master problem in just n pivot, where n is the number of nodes in the network using Limited Column Generation technique. The subproblem, column generation in each pivot, can be solved through repeatedly applying an ETKP solver.
But this heuristic has a weakness. In each column generation, the ETKP is very efficient but in the final tree partitioning process, other good cases are ignored. To improve this weakness a proper column that contain two merged other column is chosen and re-solved by ETKP process except that column. This new enhanced heuristic presents a good performance.