Research Of Topology Situational Awareness And Analysis Technology In Wireless Multi-hop Networks

There is complicated topology on wireless multi-hop networks due to various factors such as strong node mobility,nodes joining or leaving the network irregularly,attacks or random failures on the network,complicated and varied communication scenarios,etc.The traditional topological index scheme focuses on describing the static characteristics of the topology,lacking in the dynamic characteristics of wireless multi-hop topology,and lacks an efficient feature fusion mechanism.So it may even result in erroneous conclusions due to one-sided information.This paper proposes a method for situational awareness and analysis on wireless multi-hop network topology.It includes some key technologies such as construction of situation factors,situation analysis,and situation prediction on wireless multi-hop network topology.This method effectively integrates the data of wireless multi-hop topological situation,and presents a more comprehensive network topology state information,providing decision support for the optimization of wireless multi-hop network routing protocol parameters and improvement of network performance.The work of this paper is divided into three parts.The first part of this paper is to construct topological situation factors.First,based on a comprehensive analysis of the influencing factors of the topological situation in wireless multi-hop networks,an initial index set of wireless multi-hop topology topologies is constructed.Then,based on the Pearson product moment coefficient,the correlation of the initial indicator set is analyzed and simplified by our rules to construct the topological situation factor.Finally,we use k-Medoids to cluster the initial indicator set and the situational factors respectively,then calculate the Rand Index on two clustering results to verify that the situational factors still reserved higher integrity and low redundancy compare with the initial indicator set.Second,this paper presents a topological situation analysis method.The situation factor is just a collection of data,and it is inconvenient to make an intuitive interpretation of the state of the topology.Firstly,we use the k-Medoids cluster analysis and expert confirmation to obtain situation labels for the situation factor samples.Then we use the SVM classifier to analyze the connectivity situation and dynamics situation of the network topology,Then we evaluate the the accuracy rate of SVM classifiermodel based on the k-fold cross validation method.Finally,by comparing the difference in the topological dynamic analysis between the topological situation analysis and the Relative Moving Speed index,our experiments show that the topological situation analysis not only effectively fuses several index information used to describe the dynamics of the topology,but also can describe the dynamic characteristics of the network topology in real time and accurately.Third,this paper conducts a related research on topological situation prediction.Situation prediction refers to the prediction of the future trend of the topological situation,which can improve the awareness of topological status of network management personnel.The paper first presents the topological situation prediction process based on the SVM algorithm,and then designs three groups of experiments to explore the predictability of topological situation factors.The results of the first two sets of experiments show that in a specific network scenario,prediction of topological connectivity situational factors can achieve better prediction results.In the second set of experiments,the Average Neighbor Change Number index that characterizes the dynamic characteristics of the topology shows drastic changes and irregular characteristics,so the prediction effect is not good.The third experiment explores the predictability of the mean of the Average Neighbor Change Number under the condition of cyclical changes in network dynamics.The experimental results show that most of the predictions have little deviation from the actual values,and the prediction scheme is generally feasible.