Towards a System Design Discipline

点击翻译按钮获取中文摘要

ECS-9725148 Varaiya System design is a craft and a discipline. Practice of the craft requires domain knowledge. This proposal seeks to formulate a discipline of design through a graduated approach to system description frameworks, behavior, implementation and performance. The approach will build on existing work in hybrid systems, hierarchical and distributed control, and elements of cognitive science and intelligent systems. Some new theoretical and computational material will be developed while filling in gaps and exploring promising directions. The end product of this effort will be: (1) articles and a monograph describing our approach to system design; (2) software tools that aid in the creation of the design, its validation and implementation; and (3) designs that offer solutions to concrete problems in transportation and communication networks. The approach is intended for engineers and students engaged in the design of large, distributed dynamical systems. The approach will be elaborated in three passes, successively deepening theoretical understanding and examining more difficult, concrete problems of design. Pass I provides a rapid introduction to system modeling in SHIFT, a programming language for representing and simulating a dynamic network of interconnected hybrid systems. Pass II is devoted to modeling frameworks, analysis of system behavior, and real-time implementation. Pass III develops a theory of system architecture and principles for architecture design. The material for Pass I already exists, so the work here is to create a felicitous notation for hybrid systems that is mathematically familiar and syntactically convenient as a programming aid. This effort should last not more than six months. The significant body of hybrid systems theory produced in the last five years provides much material for Pass II. That material will be augmented by extending recent results on approximations and robustness of hybrid systems behavior. Also needed will be a formul ation of what is meant by real-time implementation, making use of existing work in real-time languages and operating systems. This part will draw upon research in real-time automotive control in the California PATH Program. The first version of Pass II should be complete within eighteen months. Pass III will attempt to develop a theory of system architecture as a hierarchical, distributed organization of functions necessary for perception and control. The attempt will draw upon ideas from cognitive science, distributed, hierarchical systems, and intelligent systems, and upon experience in the design of automated highways and communication networks. This part will also be concerned with the evaluation of system performance (including both logical and quantitative criteria) and show how a properly structured design assists performance evaluation. Pass III is the riskiest portion of the proposed research. It should be complete within two years. The third year of the project will be devoted to testing the theory and software tools within the context of specific applications. Art introductory graduate-level course developed alongside this work will also serve to evaluate how well the approach can be learned.