Research On Wireless Network Measurement And Optimization Based On Graph Embedding

Nowadays,from mobile payment to entertainment,mobile Internet has become a part of people's life,and the requirement of network service quality is getting higher.With the increasing number of people surfing the internet with mobile devices,more and more wireless infrastructures such as WiFi access points(APs)are densely deployed,which will cause possible wireless interference.How to improve the capacity and the performance of the network at the same time is becoming a research hotspot in wireless network.On the one hand,we need to measure the network to acquire the network parameters and understand the network performance,on the other hand,we need to optimize wireless spectrum resources allocation to reduce overall interference using techniques such as link scheduling,channel allocation and medium access control.However,because of the large scale of the network,traditional war-driving measurement and crowd-sourced wireless measurement suffer from the high cost of exhausting network traversal.The network interference optimization relies on the representation of network interference situation,but conventional conflict graph model cannot capture the dynamic of the network interference and neglects the accumulative interference from the third parties.To solve these problems,we bring graph embedding technique into the area of wireless network measurement for the first time.The advantage of graph embedding technique is that it can preserve the proximity and structure properties of the graph without exhaustive measurements of the whole network.We extend graph embedding technique to wireless network area,propose an efficient wireless network measurement approach with role-based attributed graph embedding and an adaptive wireless network optimization approach with conflict graph embedding.To deal with the high cost of wireless network measurement problem,we propose an efficient wireless network measurement approach with role-based attributed graph embedding.We firstly formulate the role-based attributed graph embedding problem,and propose a joint optimization learning algorithm to derive the vector representation while preserving the structural and attribute properties of the wireless nodes.Then we introduce the wireless network measurement approach:embedding the graph into a low-dimensional vector space based on partial sampling,and using the embedding vectors as features to infer the network performance by machine learning.We show that the proposed method benefits a wide range of wireless applications including WiFibased indoor localization and wireless SINR estimation.We conduct extensive experiments based on real wireless network measurement dataset,which show that the proposed approach can achieve decent estimation accuracy with very low sampling rate.To deal with the wireless interference optimization problem,we propose a conflict graph embedding approach to assess network interference situations by representing the wireless nodes with low-dimensional vectors while preserving their conflict relationships at the same time.The conflict graph is based on conventional conflict graph model and added RSS value to capture the accumulative interference.Specifically,we introduce a sliding-window based partial measurement strategy to capture the dynamic of conflict graph,which partially measures random links in each snapshot of the wireless network,and takes the decayed measurements of the past snapshots to construct a sampled interference graph.We conduct extensive experiments based on real wireless network datasets,which show the efficiency of the proposed approach.