Cooperation through pheromone sharing in swarm routing
Master thesis
Permanent lenke
http://hdl.handle.net/11250/261714Utgivelsesdato
2007Metadata
Vis full innførselSamlinger
Sammendrag
Traditional routing protocols build routing tables that are optimized on one parameter only, this parameter is typically hop counts. With the introduction of new requirements, brought forth by a wide range of communication intensive, real-time multimedia applications, more sophisticated routing techniques are required. However, computing routes subject to different requirements and in environments with changing traffic patterns and network topologies, is often computationally excessive and the problems are frequently NP hard. Swarm based algorithms, inspired by the foraging behavior of ants are candidates to solve such routing problems. To ensure system robustness and scalability, routing should be truly distributed and adaptive. The ac{CEAS} is an adaptive, robust and distributed routing and management system based on swarm intelligence. CEAS is performing stochastic routing with fast restoration on link failures. Previous work has shown that CEAS is robust and efficient in solving complex optimization problems such as finding primary and backup paths or simple cyclic paths (p-cycles) in networks. In all swarm systems there is a tradeoff between performance and management overhead (number of management packets). The focus in this work is on reducing the overhead in terms of management packets generated in ac{CEAS}. To achieve this, a new algorithm is proposed that applies pheromone sharing between sources going to identical destinations. Performance results from simulations show that the new CEAS system presented in this report outperforms the original CEAS in most scenarios.