Application Research Of Wavelet Neural Network Prediction Model-based Clustering Algorithm In Ad Hoc Networks

As Ad Hoc network with dynamic topology and other characteristics, the performance of initial planar structure would drop dramatically with the increase of nodes number. The main way to solve this problem is to use hierarchical structure. In Ad Hoc network, we often construct hierarchical structure through a clustering approach. Therefore, the research on clustering algorithm is essential.In comprehensive comparison with some typical clustering algorithm,we propose a location-based weighted clustering algorithm based on wavelet neural network prediction model (WNNP-LWCA).The proposed clustering algorithm gets nodes’location information through GPS,and then we do pre-clustering by virtual cluster, elects a reasonable cluster head by considering factors such as the node degree; the distance of node and virtual center; node total hold time. The location next moment of node can be predicted by the established wavelet neural network, according to the predicted result counts the predictive value of total hold time. The clustering algorithm establishes cluster with minimum combination weighted as a standard in the clustering formation stage, and in the route maintenance stage the total hold time predicted is compared to the hold time threshold, if the cluster predicted will be unstable, then the pre-restoration process is started before the link breaks off to avoid the frequent fracture of the network link. Then the cluster structure will be stable and the network performance is greatly improved.Simulation was done on ns2-based platform for comparing the properties of RLWCA with conventional WCA through two different experiments,and the results showed the good performance on the presented clustering algorithm.Then WNNP-LWCA、RLWCA and Lowest-ID are simulated and compared respectively from three different view, Simulation results show that, compared to Lowest-ID and RLWCA, WNNP-LWCA showed the good performance on different node maximum speed, pause time and number of the nodes. Expecially, the packet delivery rate and number of breaking link are advanced.