Research Clustering Mechanism In MANET Based On Emergent Algorithm

Mobile Ad hoc networks, which are multi-hop, temporary and acentric networks, consist of some mobile terminals who can move freely and have equal status with wireless communication transceiver devices. The networks can be built rapidly wherever and whenever. Mobile Ad Hoc networks will have a bright future as it can provide a convenient and flexible communication in the absence of common infrastructures and greatly expand the scope of mobile communication application.According to the differences of network topology, Mobile Ad hoc networks fall into two categories: flat structure and hierarchical structure. In the flat structure, there are no management nodes which can optimize and manage the communication resources, the cooperation algorithms are complicated, and the responses to dynamic changes of the network are quite slow. Furthermore, the network processing power is weak and control overheads are large under the circumstances of multiple users and movements. In order to reduce networks cost, offer flexible and reliable communication and increase networks expandability, generally mobile ad hoc networks adopt hierarchical structure. The networks in the hierarchical structure can be divided into cluster. Cluster is the correlative nodes set. Each cluster has one cluster head and several cluster members. Clustering algorithm affects all performance indexes of Mobile Ad Hoc network.First, this paper introduces the concept, characteristic and topology structure of mobile ad hoc networks. It summarizes merits and drawbacks of flat structure and hierarchical structure. Several typical clustering algorithms in mobile ad hoc network will be reseached, then advantages and disadvantages of various algorithms also be discussed.Next, it details the conception and characters of Emergent Algorithms(EA). Applications of EA on clustering and routing are analyzed. These above are inspirations in our work.Finally, aiming at the problems that exist during iterations in ACE, this paper introduces probability mechanism and proposes a novel probability-based clustering algorithm named PACE. With the local information of nodes candidate for cluster head probability is the highest, PACE can create a global clustering network. In the case of the same probability, the node identifier is used to arbitrate the elections of cluster heads. Meanwhile, mergences of single node clusters and failure recovery can also be achieved by the strategies above. The theoretical analysis and simulation results show that PACE has a short convergence time, can found a stable and more optimized clustering structure, and provides a capability of failure recovery.