Research On Adaptive Unattended Cluster Network Damage Recovery Method

A swarm of unmanned aerial vehicles(UAVs)is a system comprised of multiple UAVs in specific control strategies to achieve collaborative tasks.This technology has been widely applied in various fields,such as exploration,monitoring,and rescue.Under complex environments,nodes in the swarm may be invalid due to internal malfunctions or external interferences,leading to network transmission failures.It results damage to the network topology,which can in turn prevent the swarm from working properly.To address the issues caused by invalid nodes in UAV swarms,it is necessary for the swarm to have adaptive network recovery capabilities,which will enable the swarm to recover network communication in a timely manner,particularly in cases of partial node failure.The thesis focuses on the problem of adaptive damage recovery in UAV swarm networks under node failure.The main researches include the following aspects:(1)To address the problem of topology variations in a UAV swarm network caused by different failure types and scales,we propose a communication cost weight indicator based on graph theory and establishes a mathematical model for cluster network.It focuses on central node failure,critical node failures,edge node failures and different scales of node failure,aiming to describe the topology of UAV swarm networks using graph theory.Furthermore,a mapping model for topology damage in UAV swarm networks is constructed.The proposed approach provides a mathematical representation and framework for analyzing the effects of node failures on the topology of UAV swarm networks.(2)To address the problem of link instability under single-node failure in a UAV swarm network,we propose a link prediction method based on whale algorithm and grey prediction theory.Firstly,by considering the communication cost between nodes in the swarm,we utilize the whale optimization algorithm to optimize the time response function of the prediction model,aiming to enhance its generalization ability for new input data.Secondly,we design a sliding window strategy to calculate real-time link prediction based on sample data.Finally,according to the prediction results,we select an appropriate node to establish a link,allowing for network damage recovery in the event of a single node failure in the swarm.Through simulation experiments,we validate the effectiveness and rationality of the link prediction method under the scenario of a single node failure.(3)To address the problem of decreased network connectivity caused by multiple node failures,we propose a UAV link selection method based on minimum communication cost and load balancing.Firstly,we design a memory-based minimum communication cost algorithm that incorporates the minimum communication cost matrix and path vector.This algorithm enables the rapid search for missing links in the network.Then,we design a path selection algorithm based on the whale algorithm,introducing load balancing technology and considering the communication cost and load balance of the swarm network.This algorithm enables the recovery of network topology connectivity in the scenario of multiple node failures,achieving the restoration of the swarm network after damage.Finally,we conduct simulation experiments to evaluate the effectiveness of the proposed UAV link selection strategy,which demonstrate that the proposed algorithm is capable of maintaining overall cluster connectivity in the presence of a 50% failure rate.Moreover,it exhibits the ability to preserve a significantly large connected subset under scenarios of massive failures,and validates the effectiveness of the proposed approach.In summary,this paper presents network damage recovery strategies for swarm network topology with different types of node-link invalidity including single and multiple node failures.The proposed approaches include link prediction and selection methods in order to provide an adaptive damage recovery solution,which offer the foundation for analysis and research in network management and application of UAV swarms.