Dynamic Adjustment Strategy Of Collaboration Among Agents For The Software Evolution

With the rapid development of software technology, it becomes an inevitable trend that software systems have the large-scale and complex design. During the development of software systems that with large scale, it is difficult to finish the development without any help. However, some mature software systems are an important asset in the society of information. Its logic and effectiveness has been demonstrated by long-term using. So, it is an effective way to develop the large-scale software systems by using the mature software unit. And it makes the using of software resource reasonably. But how to integrate the software units to organization and how to change the collaborative relationship between software units dynamically are still need to explore.This thesis studies the model and organizational structure of Agent. Combined with the characteristics of Agent technology and MAS, based on the dynamic integrated system, we proposed a dynamic adjustment strategy of collaboration among Agents that oriented software evolution. We proposed a dynamic evolution framework model based on Agent and a process model of dynamic adjustment strategy that oriented the collaborative relationship among Agents. There are two ways to complete the adaptive adjustment. One is the globally dynamic adjustment. The other is localized dynamic adjustment. The globally dynamic adjustment of Agent collaboration is implemented by the event mechanism. The localized dynamic adjustment of Agent collaboration is reflected by the dynamic adjustment of function Agent. This thesis gives two ways to show the localized dynamic adjustments of Agent collaboration. The first way is through the self-sensing of Agent to get the states of their ability and the task load, and select the applicable function Agent by the preferred mechanism to replace or add. The second way is to use the knowledge rules of function Agent. Function Agent can update or change the ability by rebinding the capability unit of Agent. In this thesis, the adaptive evolution of the system is completed by the dynamic adjustments of the collaborative relationship among Agents. It makes a full using of autonomy of the Agents.In order to described the dynamic adjustment strategy of collaboration among Agents that proposed by this thesis in detail. We introduce the model and the organizational structure of MAS firstly. The dynamic evolution framework model and the process model of dynamic adjustment strategy are proposed. Then the key mechanisms of the strategy, such as event mechanism, Agent preferred mechanism and self-sensing of function Agent, are introduced in detail. At the end of this thesis, through a simulation system which aiming at air defense, the validity of dynamic adjustment strategy of collaboration among Agents that oriented software dynamic evolution is tested. And the experimental results are analyzed.