CAREER: Structured Design of Embedded Software

嵌入式软件无处不在：它存在于大量的日常产品和家用电器中，在汽车和飞机等系统的功能和开发成本中所占的份额越来越大。从交通到军事，从消费品到制造业，许多领域的创新越来越依赖于我们设计越来越复杂的嵌入式系统的能力。然而，嵌入式软件的复杂性和复杂性正在迅速接近当前设计能力的极限。该项目旨在开发一种正式的嵌入式软件设计方法，通过开发设计结构，特别是模块化(从组件组装系统的能力)和层次结构(将复杂组件实现为较简单组件的集合的能力)来应对复杂性。该项目从三个方向阐述。第一个项目方向侧重于确保设计中使用的组件彼此兼容的方法和工具。这有助于避免由多个组件的交互引起的错误，这些组件通常是最难预防和检测的组件之一。该项目开发了组件接口理论，用于捕获嵌入式软件组件之间交互的协议、时序和性能方面。由此产生的接口理论构成了一个扩展的类型系统，该系统不仅包含作为输入和输出传递的值，而且还包含组件的动态行为，从而能够在设计时检查组件是否以兼容的方式交互。第二个项目方向是研究从系统组件的性能和可靠性中得出系统性能和可靠性的方法。与这种模块化的分析方法并行，该项目方向将采用模块化的调试方法，其中组件被隔离分析，并使用结果来指导整个系统的模拟。最后，由于嵌入式系统经常被用作设备或物理系统的控制器，第三个项目方向建议使用多模式随机系统，这是一种以集成方式支持系统识别、控制器设计和代码生成的模型。除了学生直接参与拟议的研究外，这项建议的教育部分还包括开发两门课程：系统工程导论和实时嵌入式软件。第一门课程将向学生介绍一种结构化的、基于模型的系统设计方法；第二门课程将介绍实时嵌入式软件开发的技术和挑战。这些课程将被告知这样一种哲学，即按照组成和层次结构是设计复杂系统的关键，而模型是关于系统推理的基础。这种基于模型、面向结构的方法是计算机工程和计算机科学系普遍提供的课程中的一个创新元素。这些课程将在加州大学圣克鲁斯分校软件工程学科的计划发展中发挥关键作用。

Embedded software is ubiquitous: it is present in a vast array of everyday products and appliances, and it accounts for an increasing share of the functionality and development cost of systems such as cars and aircrafts. Innovation in many fields, from transportation to military, from consumer products to manufacturing, is increasingly dependent on our ability to design ever more sophisticated embedded systems.Yet, the sophistication and complexity of embedded software is fast approaching the limit of current design abilities. This project aims at developing a formal approach to embedded software design that copes with complexity through the exploitation of design structure, and in particular, of modularity (the ability to assemble a system from components) and hierarchy (the ability to implement a complex component as a collection of simpler components).The project is articulated in three directions. The first project direction focuses on methods and toolsfor ensuring that the components used in a design are compatible one with the other. This helps to avoid bugs stemming from the interaction of multiple components, which are often among the hardest to prevent and detect. The project develops theories of component interfaces that captures the protocol, timing, and performance aspects of the interaction among embedded software components. The resulting interface theories constitute an extended type system that encompasses not only the values passed as inputs and outputs, butalso the dynamic behavior of the components, enabling to check at design time whether the components interact in a compatible way. The second project direction investigates methods for deriving the performance and reliability of a system from that of its components. In parallel to this modular approach to analysis, this project direction will pursue a modular approach to debugging, in which the components are analyzed in isolation, and the results are used to guide the simulation of the entire system. Finally, as embedded systems are often used as controllers of a device or physical system, the third project direction proposes the use ofmulti-modal stochastic systems, a model that supports in an integrated fashion system identification, controller design, and code generation. In addition to direct student involvement in the proposed research, the educational component of this proposal consists in the development of two courses: Introduction to systems engineering, and real-time embedded software. The first course will introduce students to a structured, model-based approach to system design; the second will present the techniques and challenges of real-time embedded software development. These courses will be informed by the philosophy that structure, in terms of composition and hierarchy, is the key to the design of complex systems, and that models are the basis of reasoning about systems. This model-based, structure-oriented approach is an innovative element in the curriculum generally available in Computer Engineering and Computer Science departments. The courses will play a pivotal role in the planned growth of the Software Engineering discipline at UC Santa Cruz.