Design And Implementation Of Time-Space Uncoupled Interest Management In Complex System Distributed Simulation Platforms

Complex system is an important subject in complexity science. The research of complex system is concerned increasingly. In these research fields, the primary approach to scientific research and practice is system modeling and simulation. Complex system simulation platforms are infrastructures of complex system simulations. At present, most of the complex system simulation platforms are agent-based, which are based on CAS (Complex Adaptive System) theory. And some distributed versions of these platforms are being designed, and this trend is irresistible. Interest management is one of the key technologies of distributed simulations. It has great sense to implement interest management service in these platforms.Based on CAS theory, this thesis addresses some issues related on time-space uncoupled interest management in distributed platforms for complex system simulations. And the implement of interest management service in Advanced JCass (Advanced Java-based Complex Adaptive Systems distributed Simulation platform) is also discussed.Firstly, this thesis induces the pattern of agents' external behaviors and the abstract structure of agents by analyzing the existing representative models of agents. According to these pattern and structure, appropriate mode for agents' communication is chosen. Oriented this choice, spatially and temporally uncoupled communication is designed, which is based on the publish/subscribe paradigm and the mailbox paradigm.Secondly, based on this time-space uncoupled communication mode, data filtering mechanism is designed. Three main categories of data filtering are surveyed in this thesis. According to agent's interests, interest clustering is discussed. Especially, as region is the most widely used interest-expression, we have discussed the issues of region clustering. After these work, a multi-lever data filtering mechanism based on interest clustering is proposed, including its essential ideals, the strategy of region clustering and its basic structure.Because the speed and precision of interest matching have great effect on the performance of interest management, this thesis presents an improved sort-based matching algorithm, which based on the the runtime locality of complex system simulations. We have viewed the sort-based matching algorithm, and inspected its advantages and shortcomings. Based on the runtime locality of complex system simulations, an improved sort-based algorithm for region matching is presented, including its essential ideals, detail data structure, algorithm specification. The complexity of this algorithm is analysed, and comparative test is carried out. The test indicates that this algorithm improves the performance greatly when agents' states do not change too rapidly.Thirdly, according to the goals and missions of interest management service in Advanced JCass, we implement the interest management service in this platform. The implementation achieves these goals and missions, that is, modeling conveniently, improving simulaions performance and flexibility.