Design And Implementation Of Two Kinds Of Multi-agent Platform Integration Schemes

When we use multi-agent platform to simulate a multi-agent system, a perfect multi-agent platform should not only be able to generate agent easily, observe the interaction among agents clearly from the microscopic point of view, but also can control the simulation process in real-time from the macro point of view, and visualize simulation results are collected. However, the existing multi-agent platforms have not met these requirements. Existing multi-agent platforms can be divided into multi-agent development platform (MADP) and multi-agent simulation platform (MASP), which are functionally complementary. The former are prominent in model development. They provide the abilities, in which agents to be created easily and interaction among agents are clear, while they do not provide an intuitive and visualized graphical interface to observe the interaction among agents, and can not observe and collect the simulation results in real-time; The latter are prominent in model simulation. They can vividly simulate model, control the simulation process on the whole, observe and collect simulation results in real-time, while they do not provide function for generating agents, to say nothing of the interaction among agents. Integrating two multi-agent platforms into a new platform, combining advantages of both platforms, is extremely necessary.Admittedly, few researchers have achieved an initial realization of the platforms’ integration based on their field of study or a specific project. However, platforms’ integration requires that researchers master a large number of multi-agent platforms, choose their advantages reasonably, and achieve a large number of complex compatibility converts, which lead to the research on platforms’ integration still rarely. In view of the time characteristic differences between two kinds of multi-agent platform, two key issues must be addressed to achieve platforms’ integration. The one is giving a time concept to the agents which were developed based on MADP; the other is to let the agents developed based on MADP achieve time synchronization with the simulation environment of MASP.This paper presented a viable scheme achieving integration of two kinds of multi-agent platform effectively. The scheme changes the agents’ state in their respective life cycle on MADP, so that agents achieved the waiting in time by which the time concept was added to agents, improves the ACL (agents communication language), so that it can handle the communication behavior among agents with a time nature, to build a complete model; sums the time that the agents from generated to active based on MADP and the time that required for a single round of the game simulation as the time span of one tick in the simulation environment of MASP to achieve time synchronization of agents and MASP. Thus, the main tasks carried out as follows:1. Introduced Complex Adaptive Systems Theory to analyze the existing system, gave the existing multi-agent platforms a reasonable classification, and proposed the research idea of platform integration.2. Adhered to the idea of platform integration, the two kinds of multi-agent platform integration schemes were studied.3. Selected special cases of two kinds of multi-agent platform, staged elaborate a scheme to achieve the "Select and Add " integration of two kinds of multi-agent platform.4. Selected test model, the platform integration scheme designed in this paper was verified and analyzed.Our results provided a strong support in theory for further study of multi-agent simulation method, technically provided the ideological basis for the design and development of new multi-agent simulation platforms.