Research On Adaptive Strategy Of Dynamic Software Architecture Model

With the rapid development of computer technology and Internet technology, the software system has been applied to all trades and professions of people's lives, more and more software systems deployed on the Internet platform, making the speed of information exchange and sharing of software systems faster, functionality of the system has also been enhanced. However, there are new problems brought to software system due to the uncertainty and unpredictability of network environment, namely how to make the dynamic changing network environment flexibility of the system better, stronger robustness, scalability and more easy to become concerned about software engineering and research focus, that is how such a dynamic changing network environment flexibility of the system better, stronger robustness, scalability more easily become a hot field of research and concern of software engineering. To increase flexibility, robustness, and scalability of the system, the researchers hope that the relevant software system has the ability to customize the perception of environmental changes related to changes in the initiative to adjust its behavior according to the structure or to adapt to environmental changes that system has adaptive capability referred adaptive systems.Currently, the study of the adaptive systems includes adaptive sensing technology, technical planning and decision-making implementation of technical evolution. However, the pros and cons of adaptive systems ability direct impact on the system's flexibility, robustness and scalability, so adaptive strategy has important scientific and theoretical significance. At present, the development of adaptive strategies primarily by software designers and developers to develop based on past experience, but because of the uncertainty and unpredictability of the environment, so that the development of this policy is not strong adaptability and the intelligence is not high. Therefore, we focused on the adaptive strategy is not high intelligence, adaptability is not strong and other issues to study in decision planning theory of adaptive system.Firstly, from a global perspective and macro point of view, based on the principle of separation of concerns, establishing a framework of adaptive software model of a hierarchical based on dynamic software architecture and exploring the operational mechanism adaptive; secondly, for the problem of the current intelligence and resilience adaptive strategy is not strong, adaptive strategy is divided into pre-defined policies and non-default policy research respectively. For pre-defined policies based on ECA(Event Condition Action) adjustment mechanism, from the user's perspective and the perspective of suppliers, we present a composite based on the priority of utility theory ECA adjustment mechanism to address user preferences and multi-objective problem; for non-pre-defined policies, this paper combines reinforcement learning techniques in machine learning, machine learning reinforcement learning techniques, for non-default policy system was customized online learning, and improved learning algorithms to improve the learning speed and the convergence period, so make real-time online system enhancement strategy. Finally, the improved algorithm is applied to the robot plan the optimal path in the maze environment cases are analyzed to achieve in order to verify the correctness, feasibility and effectiveness of the algorithm, experimental results show that the improved learning algorithm improve learning speed and convergence cycle than previous learning algorithms.