Using Dead Ants to improve the robustness and adaptability of AntNet routing algorithm

This paper studies the characteristics and behavior of AntNet routing algorithm and introduces two complementary strategies to improve its adaptability and robustness particularly under unpredicted traffic conditions such as network failure or sudden burst of network traffic. Both of the proposed strategies use the knowledge of backward ants with undesirable trip times called Dead Ants to balance the two important concepts of exploration and exploitation in the algorithm. Although in AntNet routing algorithm Dead Ants are neglected and considered as algorithm overhead, our proposal uses the experience of these ants to provide a much accurate representation of the existing source–destination paths and the current traffic pattern. In the sense of routing process, gathered data of each Dead Ant is analyzed through a fuzzy inference engine to extract valuable routing information. This information is then refined according to their validity and added to the system׳s routing knowledge. In the sense of traffic monitoring, arriving Dead Ants and their delays are analyzed to detect undesirable traffic fluctuations and used as an event to trigger appropriate recovery action. To have a comprehensive performance evaluation, our proposed algorithm is simulated and compared with three different versions of AntNet routing algorithm namely: Standard AntNet; Helping Ants and FLAR. Simulations are run on four different network topologies under various traffic patterns. As simulation results show, improvements of our algorithm are apparent in both normal and challenging traffic conditions.