Software Architecture Self-adaptation Model And Intelligentizing Research

As the developments of the object-oriented technology and the component technology, the scale and the complexity of the software system have been increased. And at the same time because of the development of Internet, the environments of the software systems are much more open and change dynamically. So the software systems are harder to manage and maintenance today. In order to reduce the management and maintenance cost, and to control it from senior abstract level, research about software architecture self-adaptation arises.Using architecture as an external model to support the process of the software self-adaptation is currently an important research direction. The overall benefit from the system depicts the current configuration state. It is beneficial for system-level feature attributes to be monitored. The current study of adaptive software architecture focuses on off-line planning. Different software architecture styles have different adaptabilities to realize, that is to say, the style of the architecture has great influence on the feasibility to make the software have adaptability. So use a framework BASE to evaluate the architecture styles.Reinforcement learning theory and SARSA algorithm are proposed to realize the decision-making process, which can learn from the environment. And a frame model which contains user layer, adaptation layer and system layer is also proposed. What's more reinforcement learning based decision model was proposed.SARSA based online planning software architecture adaptive method is also proposed. The environments of the complex software systems have the inherent uncertainty, complexity and unpredictability, so off-line planning has some limitations. On-line planning methods that can automatically select the action based on the current state of the environment. This paper describes the key elements of online planning methods and process strategies. Finally a robocode instance named S ARS ABot is used as a case to fight with FireBot. The result of the experiment proves the feasibility and effectiveness of the SARSA based on-line approach.