Research On Multi-UAV Network Based On Consensus Strategy

In recent years,unmanned aerial vehicle(UAV)has been widely used in various fields of people’s life because of its small size,strong mobility and low cost.Compared with single UAV,team UAV system has significant advantages in performing cooperative tasks such as tracking,monitoring,inspection and automatic chemical plant.Among them,as an important application in the field of UAV,multi-UAVs cooperative tracking task has attracted extensive attention.In the multi-UAVs tracking network,the accurate and timely information interaction between UAVs is the key to accurately track the target.At present,more and more attention has been paid to distributed collaborative estimation methods.Multi-agent consensus strategy is one of the important methods.An effective consensus strategy can complete the tracking task through the information interaction with adjacent UAVs.However,in the multi-UAVs tracking network,because the UAV of the tracking network is moving all the time,and the channel environment is complex and changeable,the network topology will change at any time,and some existing communication links will be interrupted frequently,resulting in the decline of communication quality,which seriously affects the accuracy and convergence delay of the common strategy,so that accurate tracking cannot be realized.In the tracking network,the main factor affecting the consensus convergence delay is the connectivity of the network,and the energy efficiency of UAVs are also the most basic performance index in the process of tracking and consensus.Therefore,this paper will study the above indexes as follows:Aiming at the way of increasing or decreasing network topology edges to improve network connectivity,we introduce a relay UAV into the tracking network,establish a new network topology update method,and propose a scheme of relay UAV deployment and trajectory planning based on DQN algorithm.We divide the process of deploying relay UAV into two steps.Firstly,a deployment algorithm based on DQN is proposed to obtain the initial position of relay UAV,which optimizes the energy consumption and consensus convergence delay of tracking network.Secondly,in the process of tracking the movement of UAV,a trajectory planning algorithm based on DQN is proposed,which effectively improves the consensus success probability of multi-UAVs tracking network and reduces the interruption probability of multi-UAVs tracking network.Simulation results show that our algorithm has better performance than Q-learning algorithm and random location deployment algorithm.Aiming at the way of increasing or decreasing the network topology weight to improve the network connectivity,we propose a joint optimization scheme of multi-UAVs trajectory and power allocation based on MADQN algorithm.Due to the non-convexity of the joint optimization problem and considering that the utility of each UAV is determined according to the network environment and the behavior of other UAVs,we can model the problem as a random game problem.Aiming at the problem of high computational complexity caused by continuous action space and large state space,MADQN algorithm is proposed to solve the optimal strategy of the problem.Simulation results show that our algorithm has obvious advantages over other algorithms,and effectively reduces the consensus convergence delay and network energy consumption.In view of the specific reasons why changing the network connectivity can reduce the consensus convergence delay,we explain the network consensus problem from the perspective of graph theory,analyze the network connectivity,and give the specific reasons why the first two proposed schemes can improve the consensus convergence delay in detail.At the same time,they are simulated respectively,and the simulation results verify our analysis.To sum up,this paper mainly studies the two performance indicators of network consensus convergence delay and energy efficiency in the tracking scenario,makes the corresponding analysis from the perspective of network connectivity,and puts forward two optimization schemes respectively.The future research directions are prospected.