Testing embedded complex applications- modeling methodology and experimental environment

最近在信息获取、处理和传输方面的技术创新鼓励了集成硬件组件的复杂应用程序的开发。此类应用程序与以干扰(噪音、故障等)为特征的环境交互。这可能会在运行时导致性能下降。尽管这些嵌入式复杂应用很重要，但它们的开发方法仍然无法保证和优化所需的服务质量(Qos)，即使在设计阶段考虑了自适应和体系结构重新配置技术。主要原因是在设计和开发过程中没有考虑组件的整体包装。这尤其适用于硬件组件：实际硬件组件的成本太高，无法用于测试目的，而且硬件组件的快速技术发展使其难以使用现有的软件建模方法进行模拟。我们研究的目的是研究和开发一种方法，用于对组成复杂应用程序的软件和硬件组件进行规范和建模。我们将研究进化处理的影响，以优化这些组件的服务质量，并使它们适应环境的变化。我们的研究将不可避免地包含实验维度，因为我们模拟组件环境中的行为变化，以定义维护这些组件的服务质量的策略。为了定义我们的方法，我们将HRT-Hood等对象软件工程中成熟的方法与进化和模拟技术相结合并进行扩展。我们开发了一个通用的试验台来测试各种组件上的进化技术。然后，我们在真实的硬件组件上验证了我们的实验。为了提供更强大的自适应软件功能，我们在我们的开发框架中定义的整个复杂应用程序上测试组件结果。

Recent technological innovations in information acquisition, processing and transmission have encouraged the development of complex applications integrating hardware components. Such applications interact with environments characterized by perturbations (noises, faults, etc.) that can cause performance degradation during runtime. Despite the importance of these embedded complex applications, their development methodologies still fail to guarantee and optimize the required quality of service (QoS) even when adaptive and architecture reconfiguration techniques are considered at the design phase. The main reason is that the total package of components is not taken into account during design and development. This applies especially to hardware components: real hardware components are too costly to be used for testing purposes, and the rapid technological evolution of hardware components makes them difficult to simulate with existing software modeling methodologies. The purpose of our research is to study and develop a methodology for specification and modeling of software and hardware components that compose complex applications. We will investigate the impact of evolutionary processing to optimize the QoS of these components and make them adaptive to changes in their environment. Our study will contain inevitably experimental dimensions since we simulate behavior changes in the environments of components to define strategies that will maintain the QoS of these components. To define our methodology we combine and extend well-established methodologies in object software engineering such as HRT-HOOD, with evolutionary and simulation techniques.  We develop a generic testbed to test evolutionary techniques on various components. Then, we validate our experiments on real hardware components. To provide more powerful adaptive software capabilities, we test component results on entire complex applications defined from our development framework.