Font Size: a A A

Middleware-based adaptation evolution with reusable adaptation components

Posted on:2013-10-06Degree:Ph.DType:Dissertation
University:Universiteit Antwerpen (Belgium)Candidate:Gui, NingFull Text:PDF
GTID:1455390008982537Subject:Engineering
Abstract/Summary:
With recent development of mobile and pervasive computing, software applications are increasingly expected to dynamically adjust their behaviours according to the highly dynamic environments they are deployed in. Applications must sense the environment changes and reacting upon those changes based on their contextual knowledge. This results in the high complexity of adaptive software development, especially for the construction of adaptation behaviours.;Multiple approaches normally focus on providing support for particular set of systems and with predefined quality-of-service optimization goals. Therefore, they are costly to reuse in new systems and hard to adjust to other concerns. How to streamline the engineering of adaptation with multiple and evolving combinations of concerns remains a largely unexplored topic.;This work addresses this issue from two aspects: Firstly, Transformer, a novel adaptive framework, is proposed by extending the separation of concerns paradigm to adaptation logics design. System's global adaptation behaviour is contextually constructed with multiple reusable adaptation modules each of which embeds an adaptation strategy limited to one or more concerns. Rather than assuming these strategies to be orthogonal and thus not interfering with each other, this framework provides a systematic and customizable conflict detection policy and resolution mechanism. Secondly, a modular middleware architecture is designed to facilitate the incremental deployment of new and unforeseen adaptation modules. Software engineers are provided with the ability to add/remove/update adaptation modules during run-time. Development of adaptation modules is simplified by factoring out common adaptation mechanisms.;Design evaluation and experiment results show that our system has significant advantage over traditional approaches in light of flexibility and reusability of the adaptation modules, with little complexity and performance overhead. Moreover, it was applied in a practical case study – autonomous robots control. Experience gained from this case justified that both the framework design and modular middleware-based implementation add significant value to developers in designing and incorporating new adaptation logics.
Keywords/Search Tags:Adaptation
Related items