Kudelski, M; Pacut, A
We analyze a SWARM-based multi agent control scheme for controlling the traffic of data packets in ad-hoc networks. We consider nonstationary traffic patterns. We demonstrate how the distributed and geographically localized knowledge gathered by ant agents may improve the effectiveness of the ant learning mechanism. Our experiments indicate the improvement of adaptation capabilities of ants under dynamic topology changes and dynamic load level changes in the network.
Keywords: biometrics