Research On Anti-interference Of UAV Cluster Coordinated Airspac

Unmanned Aerial Vehicle(UAV)swarm has been widely valued for its advantages such as platform miniaturization,distributed flexible organization,low cost and strong mobility.However,the open wireless communication characteristic of UAV swarm network makes it susceptible to interference,so the anti-jamming technology of UAV network has become one of the key research directions.By using beamforming technology,the signal to noise ratio can be optimized and the interference can be suppressed effectively.In this thesis,the beamforming decision is modeled as a Markov decision process,and based on deep reinforcement learning architecture,the collaborative spatial domain interference of UAV swarm is studied.The main content and innovation of the thesis are as follows:(1)In view of the lack of signal data set in the research of UAV swarm spatial domain antijamming algorithm,the software is used to simulate the realistic scene model,and the data with real significance is simulated according to the model,and the data set is made to fit the spatial domain anti-jamming deep reinforcement learning;Based on this data set,the anti-jamming decision network of deep reinforcement learning is established,and the anti-jamming optimization of beamforming in the simulation of UAV flight path is realized.Simulation results verify the convergence and effectiveness of the proposed anti-jamming algorithm.(2)Considering that the receiver cannot know all incoming wave directions of interference signals when it performs spatial domain anti-jamming in an unknown and dynamic environment in practice,a data-assisted collaborative spatial domain anti-jamming algorithm based on UAV swarm is proposed to solve the problem of insufficient training samples due to the lack of behavioral data collection of jammers in the actual scene of deep reinforcement learning.The convergence of the proposed anti-interference algorithm and the effectiveness of data assistance are verified by simulation.(3)In order to solve the problem that there is too much feature information in the interference data set and the traditional deep learning network does not pay enough attention to the features that need to be focused on to resist interference decision-making,based on the existing network,two attention mechanisms of efficient channel attention and convolutional clock attention module are introduced respectively,and corresponding attention modules are integrated into the anti-interference deep learning architecture.The anti-interference performance of the algorithm is improved by increasing the attention to the target information and suppressing unnecessary characteristic information.