The Clustering Algorithm And Cluster Reconstruction Algorithm In AD HOC Network

In Ad Hoc network, nodes are self-organized into network without the need of relying on communication infrastructures which are set up in advance. These nodes not only have the function of ordinary mobile terminals, but also have the ability to forward packets. Compared with the traditional fixed network, Ad Hoc network has un-center, self-organizing, multi-hop characteristics. How to manage the network effectively has become a research focus of scholars both at home and abroad when the network becomes large.Ad Hoc network generally includes two structures:plane structure and hierarchical structure. Hierarchical structure divides the Ad Hoc network into several clusters. The created classful routing can reduce the control overhead of flooding broadcast and message interaction in the process of network management. So hierarchical structure can be well adapted to the need of large-scale network management and maintenance. Current clustering algorithms are trying to create and maintain a cluster structure which can cover the entire network with less computation and communication overhead, but we should develop different clustering strategy in different application scenarios. So the weighted clustering algorithm based on weighting factor optimization and the dynamic perception-based distributed clustering algorithm are proposed according to the centralized and distributed scenarios.In this paper, the main research contents and results are as follows:1) In the centralized scenario, upper management center generates clusters by collecting the information of nodes. In the traditional weighted clustering algorithm, the selection of weighting factor will produce a great influence on network load balancing. In order to get the optimal number of cluster heads and the degree of optimal load balancing, this paper optimizes the weighting factor combination according to the actual network condition and use the firefly algorithm to find the weighting factor combination which can get highest fitness value. 2) In the distributed scenario, the proposed dynamic perception-based distributed clustering algorithm can adapt to the high mobility of nodes and try to reduce the number of re-clustering. Node connectivity, electric power, speed are considered into this algorithm which can generate clusters quickly by exchanging message. In the process of cluster maintenance, the node which leaves the original cluster can choose new cluster to join or become cluster head itself. If a node’s electric power isn’t enough, its function will be transferred to the backup cluster head. This algorithm also solves the problem of performance degradation caused by too many isolated clusters.In conclusion, this proposed the weighted clustering algorithm based on weighting factor optimization and the dynamic perception-based distributed clustering algorithm can solve the weighting factor combination optimization problem in centralized scenarios and the clustering and cluster reconstruction problem in distributed scenarios.