Heuristic methods for the design of edge disjoint circuits : An application of fiber rings in telecom networks

In this thesis heuristics are studied for the design of edge disjoint circuits. These are applied to a telecommunication network design problem: the roll-out of Fiber to the Cabinet. This problem setting can be represented by a graph. A small subset of the vertices are street cabinets and one vertex is the central office. The objective is to minimize the sum of the circuit costs to connect all street cabinets to the central office, while satisfying capacity restrictions and edge disjointness within each circuit. In the studied graphs, inexpensive paths are present with available ducts, which complicate the analysis considerably. The problem is NP-hard, since it is a generalization of the NP-hard Capacitated Vehicle Routing Problem. Therefore, an insertion heuristic is developed that uses an extension of the work of Suurballe and Tarjan (1984) to construct edge disjoint shortest paths. After the construction of an initial solution, local search is used to improve the solution. In previous work on network design problems and on vehicle routing problems, this algorithm of Suurballe and Tarjan (1984) has never been used. There the focus was on exact approaches that take very long to solve for large instances or on heuristics that use routing techniques that quite often lead to high routing costs. Finally, the performance of the insertion heuristic is compared to the cluster first - route second heuristic developed by TNO. The insertion heuristic clearly shows better performance for almost all types of instances that were studied. Especially for large instances, the cluster first - route second heuristic is completely outperformed.