Roteamento em redes tolerantes a atrasos: intensificação versus exploração no processo de busca por melhores caminhos

Abstract

Designed to Delay Tolerant Networks (DTNs), the Cultural GrAnt (CGrAnt) routing protocol uses Ant Colony Optimization metaheuristic to represent the population space of a Cultural Algorithm. The CGrAnt employs distinct components knowledge in order to explore the network characteristics and improve the message forwarding: Domain, Situational, and History. Domain knowledge plays a central role in the protocol operation as it influences the History and Situational knowledge, by determining if a node must explore (through the selection of new message forwarders) or exploit (through the selection of previously found message forwarders) the search space. By using a specific metric that analyzes the local dynamics of node mobility, the Domain knowledge can set the status of the path search (exploration or exploitation). The use of this metric can induce false- positives or false-negatives when the CGrAnt protocol evaluates the quality of a node as a message forwarder. In order to mitigate the limitations of the CGrAnt’s original metric, this work proposes three alternative metrics to the Domain knowledge of the CGrAnt. The proposed metrics cover aspects of the network which are not addressed by the original metric. Results show the new proposed metrics increase the CGrAnt performance as they achieve lower message redundancy ratio for all the scenarios considered and higher message delivery ratio for two scenarios.