Research On Software Adaptive Technology For Dynamic Architecture

In the field of software engineering,with the increasing complexity of software system,and the changes of external factors it faces which are going more and more frequently daily,a self-adaptive mechanism is needed to deal with these challenges,this kind of mechanism enables software system to perceive changes of the surrounding environment in an open state,and make corresponding adjustments to ensure that the system is in its optimal state.This dissertation analyzes and studies a classic hierarchical self-adaptive framework proposed by Cheng and David.Based on this,a dynamic self-adaptive framework for centralized management is proposed.In this framework,the following improvements have been made: Firstly,in the design of knowledge base strategy,to prevent from the lack of expansibility of condition description,this dissertation introduces postfix expression to describe the condition and improve its expansibility.Secondly,in order to avoid the strategy matching that multiple executable strategies cannot implement strategy selection,utility theory is introduced to select the best of the matched multiple executable strategies.Finally,in order to reduce the high coupling degree of component connectors of the structure in the self-adaptive adjustment process,and when cross-service self-adaptive adjustment cannot be carried out under distributed conditions,this dissertation introduces a distributed management center which is used to coordinate the dynamic adjustment of the structure.In addition,a self-adaptive strategic protocol model based on intensive learning is proposed to solve the problem that the self-adaptive software based on preset strategy cannot cope with the environment change that are beyond the strategy.In this model,firstly,the running system is mapped into a combination of state and behavior by method of using software engineering which needs to be based on target and scene,then the intensive learning technique is used in the strategy planning to search and learn in the target tree scene,so as to obtain the optimal adjustment scheme,and realize the self-adaptation to the unknown environment.Then,by using UML to model the self-adaptive system,the model can only reflect the static structural relationship and dynamic transferring relationship of the self-adaptive system,but it cannot reflect the self-adaptive ability of the software.This dissertation uses an selfadaptive modeling language built by UML expansion technology,which can be used to graphically describe the self-adaptive software.Finally,referring to some excellent modeling software and combining with the self-adaptive modeling language proposed in this dissertation,a self-adaptive modeling platform Violet is designed and implemented.And with this platform,a self-adaptive UAV flight controlling system with practical application is taken as an example for modeling description.