Study On Decision-Making Method And Parameter Learning Of CNN/BN For UAV Teams On Air-to-Ground Task

Decision-making for Unmanned Aerial Vehicle(UAV)teams on air-to-ground task plays a core part in the team collaborative combat,where the method of situation assessment and tactical decision-making is the key to ensure the effectiveness of UAV teams combat.Under the integrated task environment of modern battlefield,it's become higher and higher for UAV teams to response to uncertain environment,which needs the better method of situation assessment and tactical decision-making.Combined with the process of air-to-ground task of UAV teams,this paper puts forward some new ideas and methods from three aspects: model,optimization and parameter learning to realize the rapid and correct situation assessment and tactical autonomous decision-making for UAV teams on air-to-ground task,and the main work is as following:(1)Research on regional characteristics of UAV teams on air-to-ground task.Owing to the increasing types of threats and levels of antiaircraft defense,UAV teams are facing bigger risks in their missions than before.Firstly,the feature of radar,missiles and anti-aircraft artillery is conducted,where the three are the main threats to the flight safety of UAV teams on air-to-ground task;then the threats are modeled and analyzed,based on this,the disposition feature of anti-aircraft gun missile system is described;finally,based on the distribution of the information,the uncertainty of the decision-making and the dynamics of the process,the characteristic of UAV teams on air-to-ground task is depicted.(2)Research on the procedure of UAV teams on air-to-around task.The procedure of UAV teams on air-to-around task can be seen a set which includes multi-participants,such as the manned aircraft,UAV teams and other weapon platforms,and all the members must work closely to ensure the effectiveness of UAV teams combat.Firstly,the process of combat task of UAV teams is described in detail,which can be mainly divided into six stages: planning,cruising,penetrating,assaulting,merging and exiting;then,the characteristic of situation assessment and tactical decision-making for UAV teams is analyzed and modeled,and finally the structure of UAV teams autonomous tactical decision-making is put forward.(3)Research on situation assessment method of UAV teams on air-to-ground task.UAV teams will face the joint attack of anti-aircraft gun missile system in real-time,which can not blindly pass through the threat aera,and in such case,effective integration of the probe information from multiple sources is needed to correctly assess the situation characteristics.Primarily,the deduction platform STAGE is used for simulating and establishing the situation assessment dataset,where a way of visualizing situation is proposed;then the CNN is used to design the situation assessment model which can be suitable for the research problem,and the experiment proves its validity.(4)Research on the tactical autonomous decision-making method of UAV teams on air-to-ground task.The tactical decision-making level can directly affect the flight safety of UAV teams and the implementation of scheduled mission in the process of penetration.In the first place,based on the basic theory of probability theory,the BN and its parameters learning characteristics are analyzed;then a BN parameter algorithm is proposed to solve the BN parameter learning problem under scarce data,which integrates qualitative domain knowledge and evolution strategy,where the results show that the proposed algorithm can effectively improve the accuracy of BN parameters under scarce data;finally,the proposed algorithm is applied to the autonomous decision-making of UAV teams on air-to-ground task.