Research On Agent-Based Modeling And Its Application To Joint Operation In The Information Age

Warfare simulation is an important way to study wars, improve battle effectiveness and strengthen national defense. Especially in the Information Age, pre-research of warfare via simulation may provide the foundation for the real operational capability in the future.The modelling of forces that describe the combat entity in the battlefield is among the key technical problems in the warfare simulation domain. It is a challenging aspect of study because human is involved. Surrounding the construction of force models for the tactical joint operation simulation based on the agent theory, this dissertation studies the interrelated theory and methods from three aspects, viz. fundamental theory, system implemtation and application.Joint operation in the Information Age is totally different from the mechanization war at the points of engagement pattern and force composition. First of all, the dissertation analyses the concepts and characteristics of multi-agent systems and discusses the features and requirements of joint operation generally. Then an agent architecture named A~2FC is proposed to play the tactical joint operation simulation. The architecture is constructed according to the principle of classifying knowledge and modularizing function. It comprises domain knowledge, models and several fuction modules, including collaboration module, decision module, scheduling module, etc. This architecture describes the frame of agent, which provides the foundation of system design and implementation.Knowledge is the basis of intelligent behaviour. A set of standardization of system and methods is needed because the military domain knowledge is extremely complex. By the purpose and abstract degree, military domain knowledge is divided into two levels: ontology level and resource level. Domain ontology defines the general concepts and their relations in military domain, which provides common meanings for the communication among Agents. Resource level defines all kinds of knowledge elements involved in warfare simulation, including entities, equipments, actions, etc. The dissertation introduces the principles of constructing domain ontology and puts forward a ontology development method that consists of four steps, viz. information acquisition, ontology capture, ontology formalization and ontology evaluation. The process provides a sort of high level and operational approach for the construction of military ontology. In addition, the dissertation selects a kind of describing method for each kind of knowledge element.Decision model is the radical point of agent's intelligence. In the past the decision model basically bases on the expected utility theory. It is a purely rational theory. But human is not always totally rational. After analyzing the features of decision-making, the dissertation introduces the prospect theory to agent decision simulation to make it accord with human's decision pattern.The dissertation analyzes the concept of self- synchronization detailedly firstly. Basing on the above analysis, a collaboration system for autonomic cooperation among agents is developed, which comprises of two parts: collaboration knowledge and collaboration process. Collaboration knowledge includes terms, convention and rules that is essential to the successful implementation of collaboration. Collaboration process is defined as a method that guide the collaboration to advance successfully. Self-synchronization will come true if all participant agents follow this model. The communication between agents is the basis of multi-agent collaboration. The intentions of commanders need to be exchanged even if the physical information of battlefield can be shared through fundamental information grid. The dissertation extends the FIPA ACL aiming at joint operation.Finally, the dissertation designs an agent simulation platform and develops an application to verify the practicability of above-mentioned theory. The platform is composed of four subsystems, viz. ontology-formation, knowledge-definition, model-generation and running- infrastructure. The dissertation studies the structure of agent model deeply and discusses the constructing steps of simulation system. A scenario is made up in the context of Iran nuclear crisis. Using this script, the dissertation validates the integrated development process of simulation system and contrasts the centralized command pattern with self-synchronization in the response time. The conclusion that self-synchronization can increase the engagement speed is educed.