Robustness Analysis Of Flocking-based Swarm Algorithm For UAV Cooperative Formation

With the rapid development of the unmanned aerial vehicle(UAV)industry in recent years,UAV has become an indispensable method in dangerous operations,high-altitude reconnaissance,and material transportation.Due to the increasing difficulty of tasks,it is increasingly difficult for a single drone to meet the needs of complex tasks.Compared with a single UAV,UAV clusters have doubled their carrying capacity,detection range,combat efficiency,and survivability.Therefore,clustering has gradually become the mainstream development direction of UAVs.As the control core of the UAV cluster,the UAV cluster algorithm is the key to the cluster attitude adjustment,tactical control,and coordinated decision-making.Therefore,the research on the robustness of the cluster algorithm is of great significance for ensuring the stable and efficient operation of the cluster.At present,some research results have been achieved on the robustness of UAV cluster algorithms,but there are still some shortcomings: the existing robustness evaluation of UAV cluster algorithms is mostly done by imposing on the algorithm.different kinds of attacks,such as deleting some drones in the cluster,slowing down the speed of drones,or modifying the parameters of the cluster algorithm,change the operating environment and mechanism of the cluster algorithm,so as to observe the results and stability of the operation.However,the interference in the past work mainly focused on external attacks on the cluster or sudden abnormalities in the cluster,ignoring the vulnerabilities of the network communication protocol in the cluster environment,and lack of protection of the network and the network caused by the communication protocol vulnerabilities in the normal operation state.Research on data disturbance.Aiming at the above shortcomings,this thesis studies the vulnerabilities in the Ad hoc On-demand Distance Vector Routing(AODV),one of the cluster communication protocols,and proposes a robust evaluation of the bird swarm algorithm based on deep Q-network(DQN).The main research contents and contributions of this thesis are as follows:(1)Discovered and analyzed the loopholes in the AODV protocol.Conducted research on the current mainstream AODV communication protocol,discovered the loopholes of the protocol in the drone cluster environment,analyzed the network and data disturbances in the cluster caused by the loopholes,and discussed its impact range using the path loss formula,And then explored the impact of the disturbance on unmanned clusters through experiments.(2)The robustness evaluation model of bird swarm algorithm based on DQN is realized.According to the impact of the protocol vulnerabilities on the drone cluster,by constructing a state space for the drone cluster environment,the DQN algorithm is used to realize the robustness evaluation model for the bird swarm algorithm,and the optimal bird swarm algorithm is trained and obtained The attack strategy further supports the modification of the robustness of the bird colony algorithm parameters.This thesis is based on the network delay,data loss and data error disturbance caused by the AODV protocol vulnerability in the cluster environment.In the Gazebo simulation environment,it is found through experiments that the arrival time of the UAV cluster will be delayed by up to 49.25% under the influence of the above disturbance.Subsequently,a better attack strategy was obtained through DQN model training.Based on this,the parameters of the bird swarm algorithm were revised,and it was found through experiments that after the revision,the arrival time delay of the drone cluster decreased by 14.09% on average,and the highest drop was 37.5%,has accomplished the research goal of this thesis well.