Ad-hoc On-Demand Distance Vector Routing
WMCSA '99 Proceedings of the Second IEEE Workshop on Mobile Computer Systems and Applications
Combining Multiple Weak Clusterings
ICDM '03 Proceedings of the Third IEEE International Conference on Data Mining
| Hi-index | 0.01 |
During the lifetime of a wireless ad hoc network using the AODV routing protocol, nodes arc the sources and destinations of routing traffic. The routing traffic typically consists of routing requests, routing replies or routing errors. The routing events (requests, replies or errors) witnessed by a node could constitute an important source of information to improve the nodes routing decisions. In this work we resort to behavioral logic to equip each node with a Routing Learning Agent (RLA) to process the routing traffic in order to determine and maintain for each node involved a score (called PRS: Pairwise Routing Similarity) in the interval [0, 1]. The occurrence of a routing event at a particular node triggers the associated agent to update the scores of the involved nodes: Route replies/errors contribute to higher/lower scores. A node considers itself in the same cluster as other nodes for which it computed a score ≥0.5. A node then selectively floods received route requests to only nodes of its cluster. The nodes self clustering is used to improve routing efficiency. Simulation results using AODV show that equipping nodes with self clustering capability leads to noticeable improvement with respect to packet delivery ratio (12.9%), throughput (11.1%) and overhead (13.9%).