Cross-Entropy Guided Ant-Like Agents Finding Cyclic Paths in Scarcely Meshed Networks

Finding paths in networks is a well exercised activity both in theory and practice but still remains a challenge when the search domain is a dynamic communication network environment with changing traffic patterns and network topology. To enforce dependability in such network environments new routing techniques are called upon. In this paper we describe a distributed algorithm capable of finding cyclic paths in scarcely meshed networks using ant-like agents. Cyclic paths are especially interesting in the context of protection switching, and scarce meshing is typical in real world telecommunication networks. Two new next-node-selection strategies for the ant-like agents are introduced to better handle low degrees of meshing. Performance results from Monte Carlo Simulations of systems implementing the strategies are presented indicating a promising behavior of the second strategy.