Solving the Steiner Two-Node-Survivable Network Problem

In the design of optical fibre networks, a commonly applied requirement is to ensure the existence of at least two node-disjoint-paths between pairs of distinguished nodes. The problem of finding a topology verifying this restriction is known as the Steiner Two-Node-Survivable Network Problem (denoted by STNSNP), an NP-hard problem. This paper introduces a Greedy Randomised Adaptive Search Procedure (GRASP) for designing low-cost topologies for the STNSNP model. The heuristic was tested over a large problem set containing heterogeneous topologies with different characteristics, including instances with hundreds of nodes. The numerical results were highly satisfactory, accomplishing in all cases good quality local-optimal solutions.