The Research On Cluster And Cooperation In Wireless Ad Hoc Network

Along with the rapid development in wireless mobile communication, wireless Ad Hoc network being a new multi-hops and self-organized network is becoming the research focus. And the network's flexible features bring the new challenge on networking. In order to promote the network performance, network is divided into clusters. Stable and efficient clustering structure as an efficient networking method of promoting network performance is an important research subject in Ad Hoc network. In addition, the Ad Hoc network needs the neighbor nodes' cooperation to forward packets, but it is contrary to the personal benefit, such as power. As civilian application of Ad Hoc network, the cooperation research is concerned.First, the Ad Hoc network's feature, network topology and MAC access mode are introduced. Then, the typical clustering algorithm in clustering structure and cooperation model in cooperation research are introduced.Second, a dual-cluster-head clustering algorithm is presented. After the cluster-head is formed, the reserved cluster-head is selected. The reserved cluster-head is substitute of cluster-head. When the cluster-head is not fit its job, the reserved cluster-head become the new cluster-head. Simulation shows the algorithm reduces the mobility influences to cluster structure and promotes the cluster stability.Then after analyzing the clustering, a method of optimizing gateway dominating set by comparing the exclusive id of each node is presented. The optimization method can ensure only one gateway between two intersected clusters and only one pair gateways between two adjacent clusters. Simulation result shows, the method can save the broadcast packets and reduce broadcast delay effectively in the case of guaranteeing the network connectivity.Finally, a cooperation monitoring model basing on discrete Hopfield neural networks is presented. The model works in cluster unit and monitors the backbone nodes. The analytic hierarchy process is presented to obtain node action index weight. All neurons in the neural network are divided into groups according to action index weight. The neurons numbers in each group is proportional to the evaluation index weight. Then the discrete Hopfield neural network with the memory of action class and relevant treatment can monitor the node's cooperation. The example calculation shows the model's feasibility.