Reserch On Air-combat Decision Based On Dynamic Bayesian Network

In modern air combat,air-based UAV need to have autonomous air-combat capability.Therefore,the corresponding intelligent decision-making system needs to be used as a support to improve its autonomous decision-making ability,battlefield environment adaptability and fault tolerance.In order to solve the autonomous air combat decision-making problem of UAVs,this paper conducts research on air combat decision-making methods based on dynamic Bayesian networks.The UAV can take effective space-occupying position to quickly constitute the attack conditions under the guidance of flexible maneuver strategies.The paper analyzes the traditional expert system,differential strategy,dynamic programming,matrix strategy,impact diagram and trajectory prediction and other air combat decision-making methods.According to the current research status,the main reasons affecting the applicability of air combat decision-making methods are derived.On the support of analyzing tradition air-combat decision-making methods,the air combat decision-making method based on dynamic Bayesian network is put forward and engineering practice is explored.Bayesian networks have the ability to model decision-making based on expert experience and objective data.Therefore,the modeling process can make full use of the air-combat priori knowledge.After analyzing the battlefield situation factors that affect the result of air combat,the definition of network evidence nodes,which are used to perceive the battlefield environment,is completed.On the basis of evidence nodes definition,the logical relations among evidence nodes are analyzed,and the definition of network intermediate nodes is completed to extract the features of battlefield environment.Finally,according to the requirement of space-occupying in air-combat decision-making process,the executable maneuvers of UAV are determined and the decision node definition is completed.The decision-making network nodes are connected in the form of a directed arc.Their connection relationships are determined by the corresponding conditional probability table.The initial network parameters are set by air-combat priori experience to fit the pilot's decision-making mentality in air combat process.The air combat decision-making network is a standard Markov network.Referring to the Markov network inference process,a filter inference algorithm is designed to solve the network model.In order to ensure the objectivity of the decision-making process,the decision-making network should fit the statistic law of air combat results.Therefore,the parameter learning method based on Bayesian estimation theory is further studied.Through parameter learning,the network parameters are continuously optimized to further improve the network performance.After completing the method demonstration,a simulation experiment platform is constructed,and the method is verified through simulation experiments to reach a conclusion.Simulation results validate the effectiveness of the air combat decision-making methodbased on dynamic Bayesian network,which indicates that this method can effectively overcome the shortcomings of the traditional air combat decision-making method and has excellent engineering application value.The decision-making network model possesses strong adaptability to the battlefield environment,possesses error information processing capability,and can output flexible and reliable maneuver strategies in air-combat process.Besides,the maneuver strategy solving process does not rely on the aircraft dynamics model,making the decision-making network has strong portability.In addition,the network solving process overcomes the dimensionality disaster and can excellently satisfy the real-time requirements of air combat.