A Novel Weight-Based Adaptive Clustering Algorithm In MANET And Its Performance Simulations

The mobile ad-hoc network (MANET) is a multihop and wireless network which is dynamically produced without any fixed basic station or centralized management and which is extensively used. The allocation of resources can be achieved more easily and each node can obtain the basic bandwidth required by other nodes directly or indirectly through base stations in the cellular networks with central nodes. If a MANET network is divided into hierarchical instructure based on clusters, the approach used in cellular network can be extended to MANET. So it can improve the performance of MANET on condition that network is scaled into clusters by clustering algorithm ,thus more and more ad-hoc networks apply architecture of grade.The paper analyzes and compares many traditional clustering algorithms . By referring to the natures of the existing algorithms, a novel adaptive weighted clustering algorithm in MANET is proposed on the basis of WCA and AOW algorithm.And its validity and improvement in performances are also proved through the tests of simulation. Meanwhile, the paper focuses on studies about weight factors, analyzes the subjective and objective states related to calculating weight, and by integrating them introduces a relevant approach to calculate weight. Based on the theory of NAWCA ,a model for classifying is created by using Self-Organization Feature Map(SOM) neaural network.. After entering data of testing nodes,the model based on the algorithm is tested and a better classifying result of clustering is therefore achieved.The innovations of this thesis can be summarized into three points. Firstly, the average relative velocity is introducd into a novel adptive weighted clustering algorithm as one important parameter of weight, then it increases the stability and self-adaptability of cluster head.Secondly, a new approach to calculating weight is suggested by integrating subjective and objective factors.It is verified by comparison with other approaches to selecting weight .Thus the velocity of weight responding to the changes of network topology is increased .Finally, using a SOM neural network to create a classifying model enables every node to learn to identify by itself the role in MANET.