A study of multi agent based resource search algorithms

Abstract

Resource sharing in a networked environment has been an evolved field of research in both industry and academia. In a dynamic, resource constrained environment, resource discovery plays a vital role in the resource sharing process. A distributed multi agent model for a dynamic, heterogeneous network has been proposed in this paper. Furthermore, three resource search algorithms have been investigated, namely: Random Walk, Directed Search and Directed Search with Self-Organisation. An approximate matching mechanism between resources and tasks descriptions has been proposed as a solution for the trade-off problem between the accuracy and speed. As performing random search might lead to traverse the whole network and increase the failing ratio, we have introduced a heuristic directed search to overcome this problem. Based on the agents' closeness, the Agents re-organise themselves by creating an overlay network of agents which are holding semantically close resources and then direct the search message accordingly. The proposed solution has been implemented in a simulated environment and simulation results have shown a significant improvement in the search time (calculated as a number of hops).