| The cognitive behavior modeling of Computer-Generated Forces(CGF)is one of the key technologies in combat simulation,with the aim of accurately reflecting the influence of battlefield perception,situation awareness,planning and decision-making of human cognitive behaviors on wars and improving the fidelity and credibility of the systems.As one of the basic tasks,behavior modeling of spatial cognition includes modeling human capability of spatial representation and reasoning,focusing on the representation of battlefield space,situation comprehension and prediction,path planning for battlefield maneuver.However,it is difficult for current approaches to meet the practical requirements of building combat simulation systems,mainly because(1)quantitative space representations cannot describe the hierarchical and abstract formation of spatial memory of human beings,thereby hard to support flexible and efficient situation representation and path planning in large-scale terrains?(2)existing situation modeling methods strike a poor balance between the prediction accuracy and interpretability of decision-making rules?(3)path planning algorithms have not exploit the terrain topologies well and cannot emulate the human thought of hierarchical planning,thus running inefficiently and hard to satisfy the real-time needs of systems.To tackle these problems,improving the cognitive rationality of CGF spatial cognitive behavior modeling and ensuring the operating efficiency of the systems,this paper mainly includes the following contributions and innovations:(1)A framework of spatial cognitive behavior modeling based on cognitive architectures is designed to be used in combat simulation.Although the behavior modeling of spatial cognition is supportive for studying situation awareness,collaboration,task planning and tactical actions of CGF,existing researches on combat simulation is short of design of a behavior modeling framework for spatial cognition of CGF,resulting in fuzzy functional requirements,unclear logical connections and poor support of technical routes for downstream applications.Aiming at these problems,we firstly analyze the role of spatial cognitive behavior modeling in combat simulation systems,draw on important concepts and principles of spatial cognition in cognitive science,innovatively propose memory based cognitive architecture of spatial topological representation and situation representation and reasoning,and design a behavior modeling framework of spatial cognition based on the OODA cycle.(2)A method of representing spatial topology based on episodic memory modeling is proposed.Existing methods of representing the battlefield space in simulation systems mostly are uniform discretizations of data like geographical information.This type of models cannot represent the spatial structure and guide downstream algorithms for spatial reasoning,and make downstream computation inefficient due to large state space.This paper investigates the works from cognitive science on the morphology and formation of spatial memory,proposes to use self-organizing neural network ART to model the episodic memory of spatial exploration,extracting the locations and visited order of landmarks from continuous perception.We also model the knowledge transfer from episodic memory to spatial memory,and thus automatically building the entire topological model by a online learning manner.Results of simulation experiments show that the proposed method can build much sparser spatial representation model compared with the state-ofthe-art method,which effectively reflects the guidance of salient landmarks for navigation.(3)A situation representation and prediction method based on semantic memory modeling is proposed.Structurally sparse but adequately expressed spatial models are the basis and key to ensure their computational efficiency and scalability in large-scale terrains.In addition,the modeling process of the existing situational awareness method is highly dependent on the knowledge of domain experts,and it is difficult to provide a compact and explainable behavior model of observed CGFs,so they have very limited reliability.Aiming at these limitations,inspired by the research on representation of space-related semantic knowledge,this paper proposes a spatial situation semantic memory modeling method based on ART network to self-organized learning “if-then” form of state-target association rules from historical observation data.Experimental results on a large scale road network show that the proposed method can not only produce a compact rule set readable in natural language,but also maintain the prediction accuracy comparable to the existing methods under the conditions of uncertain behavior of observation CGF and possible loss of observation information.(4)A real-time path planning algorithm based on the topological structure is proposed.Path planning is the most common spatial reasoning task in combat simulation systems,which occupies more computing resources than other CGF behavior models due to its high frequency to be used and computational complexity.Therefore,computational efficiency is an important metric to measure path planning algorithm for multi-entity and large-scale combat simulation system.Given that human beings efficiently make plans based on hierarchical spatial representation model,this paper analyzes the topological structure to reduce the cost of time and space of the map preprocessing,designs new edge pruning techniques based on symmetry breaking and proves the completeness and optimality-preserving properties of these ingredients.Experimental results based on the commonly recognized benchmarks show that the proposed method is a new algorithm on the Pareto frontier in the field of path planning.The mentioned works lay a good foundation for solving the important problems related to spatial cognition in the OODA decision loop of CGF,effectively integrates the principles of cognitive science and computer modeling methods in theory,and provides a new means for improving the fidelity and rationality of CGF behavior.Moreover,they provide a strong support to solve the bottleneck problem of real-time operation of combat simulation systems,and improve the interpretability of CGF cognitive behavior model.This paper finally summarizes some interesting and meaningful directions worth studying on in the future. |
PDF Full Text Request