Decentralised routing strategies in sparsely connected circuit switched networks

To maintain a competitive edge it was considered essential for Dowty to investigate decentralised algorithms and in particular more advanced AI based algorithms using learning. The paper first describes the main types of decentralised routing algorithms selected for comparative study. In particular direct, fixed alternative, random routing, dynamic alternative routing and learning automata algorithms are considered. Two particular circuit switched networks were proposed by Dowty for the routing investigations and a parallel-C based simulator was used to build the appropriate models for the experimental investigations. Detailed experimental results are provided and observations made on the comparative merits of the various routing algorithms