Research On Modeling Of CGF-oriented Strategic Intention Recognition

Computer Generated Forces(CGF)is one of the key cutting-edge technologies in the field of combat simulation.Intention recognition behavior is an important research direction of CGF cognitive behavior modeling,which can effectively solve the problems of fixed and predictable CGF behavior mode,insufficient confrontation and coordination ability,low situation analysis and processing level in the existing simulation system.Strategic intention recognition behavior modeling is to consider the generalization of intention recognition behavior under the condition of battlefield confrontation and cooperation,which helps CGF to more strategically identify enemy and friend combat intentions.Research on CGF-oriented strategic intention recognition theory is of great significance for improving the CGF situational awareness and intelligent decision-making and humancomputer interaction based on this.The main contributions of the paper are as follows:(1)Construct a strategic intention recognition behavior modeling framework around dynamic recognition and inference in hypothesis spaceBy analyzing the modeling requirements and cognitive behavior foundation of CGF strategic intention recognition behavior in combat simulation system,the role and influencing factors of strategic intention recognition in situation awareness are clarified,and the strategic intention recognition behavior in cognitive behavior is proposed.The two types of implementations on the cognitive architecture construct a strategic intention recognition behavior modeling framework around dynamic recognition and inference in the hypothesis space,and summariz the information interaction with other cognitive behaviors.(2)A non-adversarial and non-cooperative intention recognition method based on Markov decision process is proposed,which can be effectively used in dynamic network interdiction decision problemsAiming at the comprehensive application of intention recognition and decision-making and other cognitive behaviors,a non-adversarial and non-cooperative intention recognition method based on Markov decision process is proposed.The particle filtering method is used to achieve efficient inference,and heuristic and inverse reinforcement learning methods modeling the observed agent’s decision-making behavior are used.Based on the intention recognition method to solve the problem of dynamic network interdiction problem,a blocking resource allocation method based on subjective confidence and a reconstruction method of dynamic network blocking model are proposed.The experimental results verify the effectiveness of the above method and improve the quality of recognition and decision-making.(3)A deceptive path planning method based on mixed integer programming is proposed,which solves the problem of deception behavior modeling in intention recognitionAiming at the problem that the existing intention recognition methods lack deceptive behavior representation and reasoning ability,taking the path planning problem as the application background,a deceptive path planning method based on mixed integer programming is proposed,which better solves the modeling of deceptive behavior.In particular,for the problem of computational efficiency on large terrain,combined with the spatial representation method based on sub-goal graph to improve the existing methods,the experimental results prove the scalability and effectiveness of the method.(4)A strategic intention recognition modeling method using mixed integer programming is proposed,which can be effectively used to model unilateral control and two-party game problems under adversarial conditions.Aiming at the problem of strategic intention recognition behavior modeling under adversarial conditions,a strategy-based intention recognition modeling method using mixed integer programming is proposed,which can be effectively used for unilateral identification control and two-party game in network interdiction problems under adversarial conditions.The unilateral identification control includes an intention umbiguity deduction model for the recognizer and an intention umbiguity improvement model for the recognized person,and the two-party recognition game is described based on double-layer mixed integer programming.Experimental results prove the effectiveness of the proposed method in controlling the recognition process and realizing the game between the two parties.(5)A hierarchical collaborative behavior model and a strategic intention recognition method based on a time constraint model are proposed to solve the problem of teammate intention recognition and collaboration under partially observable conditions.Aiming at the problem of strategic intention recognition behavior modeling under cooperative conditions,a hierarchical collaborative behavior model based on intention recognition and a strategic intention recognition algorithm based on time constraint model are proposed.Analyze the conversion relationships between the behavior tree and those of And-Or tree,Hierarchical Task Network,and formal grammar.The paper solves the problem of teammates’ intention recognition and collaboration in partially observable environment.Finally,the thesis summarizes the research work in depth,and gives the research problems and theoretical methods that need to focus on in the future.