Research On MANET Dynamic Clustering Management Mechanism Based On Intelligent Prediction

With the advantages of strong mobility,low maintenance costs,and strong survivability,Mobile Ad Hoc Network(MANET)has become a research hotspot in recent years.MANET is a system composed of a group of multiple nodes represented by UAVs.Each node is connected based on wireless channels,with basic functions like communication,computing,storage,and routing,forming a complex communication system to apply in fields such as emergency communication,public transportation,disaster reconstruction,and military uses.Due to the mobility of node,the topology of the network is constantly changing.To some extent,this has brought flexible deployment characteristics and rapid adaptability to new scenarios to MANET,but it is also the difficulty of the networking mechanism and network algorithms of MANET;On the other hand,the strong correlation between MANET network architecture and practical applications also provides a reference perspective for studying its topology change characteristics and network architecture.Research has shown that MANET networks using cluster networking architecture have better scalability and high availability.In practical applications,MANET often appears as a cluster structure.This paper will focus on the research of node mobility characteristics and cluster network management in MANET networks,and try to solve problem like node mobility and link prediction,gateway selection and handover,and formation control.The main contents are as follows:The first part discusses the problem of node mobility and link prediction in clustered MANET.According to the characteristics of cluster networking,this part first investigates node group mobility models,which describe how nodes within a cluster move based on reference points such as cluster heads.Subsequently,with these mobility models,this part proposes a node mobility prediction model based on Transformer.This algorithm performs long-term series prediction of node mobility information,and completes link birth/death prediction between nodes with the mobility prediction result.In addition,in order to solve the pre-training dataset problem of Transformer,this paper uses GAN to complete the expansion of the training dataset.The training and prediction architecture can well complete the extraction of node movement characteristics,which is more accurate in predicting compared to traditional prediction models.The second part discusses the problem of gateway selection in clustered MANET.In a clustered network,gateway nodes need to undertake a series of tasks such as cross cluster traffic forwarding,network topology maintenance,and network performance monitoring.Therefore,the physical channel quality and packet congestion degree of gateway nodes determine the quality of inter cluster communication throughout the network.Taking cross cluster channel quality and low transmission delay as optimization indicators,this paper proposes a cross cluster gateway pair selection algorithm.The algorithm models the gateway selection process as a Markov Decision Process(MDP),and solves this MDP problem based on Q-Learning.Considering the mobility of gateway nodes,this part introduces the movement prediction information of candidate gateways and adjacent nodes as model modifications.Experiments show that this gateway selection mechanism has a good performance improvement on network performance.At the same time,the link prediction revising makes the algorithm have good adaptability to network topology changes.The third part discusses the cooperative formation control problem of MANET.Formation control is an important factor affecting the performance of MANET.On the one hand,collaborative formation is the foundation of physical performance optimization such as energy consumption control,and on the other hand,it is also an important way to form an efficient topology and optimize upper layer network parameters.This part focuses on the process of nodes moving from an unorganized and unclustered state to forming a reliable formation and cluster,and proposes an efficient collaborative algorithm that can complete the formation within a specified time period with lower formation flight energy consumption as the optimization index.Similarly,this scheme introduces the mobility prediction results of the first part as an optimization for collision avoidance between nodes and reducing the length of the routing path.Experiments show that the proposed formation cooperative control algorithm can avoid collisions and form a networked formation with a high success rate within a specified time,while achieving performance optimization at a lower energy cost and additional path consumption.