Research On Multi-function Radar Working Mode Recognition Based On Multi-agent Predictive State Representation

Electronic reconnaissance,an important part of electronic countermeasures,can provide necessary intelligence information for combat tasks such as perceiving battlefield situation,identifying targets,and estimating threat degree.Indeed,radar working mode recognition is the basis of electronic reconnaissance.With the rapid development of information technology,Multi-Function Radar(MFR)is gradually replacing traditional radar and has been widely used due to its rich functions,strong reliability,and complex signal patterns.As a result,the agent-detected information is an incomplete description of the dynamic MFR behavior pattern when a single unmanned reconnaissance aircraft is used as a single agent to perform identification tasks.In view of the limitations of single-agent identification,this dissertation is oriented to the scenario of swarm UAV electronic countermeasures,and a multi-agent system is constructed by multiple unmanned reconnaissance aircraft,and the integrity of reconnaissance information is improved through information fusion between agents.However,there are the following difficulties in the fusion of information among multiple agents.Firstly,the perception contents between some unmanned reconnaissance aircraft are not correlated,which ensures the consistency of the opposed targets.Secondly,the amount of information detected by each unmanned reconnaissance aircraft varies,and there is a situation in which some unmanned reconnaissance aircraft detect less information and have weak information complementarity with others.In response to the above issues,this dissertation takes advantage of gray correlation analysis,graph filter and predictive state representation(PSR)model to construct a multi-agent PSR model recognition framework to conduct research on MFR working mode recognition.The details of this dissertation are as follows:(1)A method for MFR working pattern recognition based on graph filter and multi-agent PSR model is proposed,which addresses the issues of inconsistent perception of targets,low amount of interception information and weak complementarity among some reconnaissance agents when multiple unmanned reconnaissance aircraft jointly intercept information.the problem of target consistency is solved firstly by gray correlation analysis,which ensures the validity of information during fusion processing.On this basis,graph filters are constructed to realize information sharing and fusion among multiple unmanned reconnaissance aircraft to improve the integrity of information.After that,the system dynamic matrix is constructed to describe the transfer rule of MFR signals,and the PSR model is established by processing the dynamic matrix with the core search algorithm.Then,by utilizing the PSR model,the single-step prediction probability distribution is calculated for each working mode.Finally,to realize the working mode recognition of MFR,the grid filtering algorithm is used to track the dynamic changes of the probability distribution according to the transition probability between the working modes.The results of simulation experiments have proved that in the case of the same error radar,the method mentioned in this article has increased by about 2% compared to the traditional method.(2)The transfer probabilities between the working modes are not easily obtained in the actual adversarial environment.Consequently,this dissertation proposes a radar working mode identification method based on the PSR-Elman network model.The key information processed by the PSR learning algorithm is used as the input of the Elman network model,and the actual working mode is used as the output.Taking advantage of the network’s advantages in fitting dynamic systems,the mapping relationship between MFR signal observations and working modes is constructed to realize the identification of the MFR working mode.The simulation experimental results show the effectiveness of the proposed algorithm.