A comprehensive methodology for building hybrid models of physical systems

摘要

This paper describes a comprehensive and systematic framework for building mixed continuous/discrete, i.e., hybrid physical system models. Hybrid models are a natural representation for embedded systems (physical systems with digital controllers) and for complex physical systems whose behavior is simplified by introducing discrete transitions to replace fast, often nonlinear dynamics. In this paper we focus on two classes of abstraction mechanisms, viz., time scale and parameter abstractions, discuss their impact on building hybrid models, and then derive the transition semantics required to ensure that the derived models are consistent with physical system principles. The transition semantics are incorporated into a formal model representation language, which is used to derive a computational architecture for hybrid systems based on hybrid automata. This architecture forms the basis for a variety of hybrid simulation, analysis, and verification algorithms. A complex example of a colliding rod system demonstrates the application of our modeling framework. The divergence of time and behavior analysis principles are applied to ensure that physical principles are not violated in the definition of the discrete transition model. The overall goal is to use this framework as a basis for developing systematic compositional modeling and analysis schemes for hybrid modeling of physical systems. Preliminary attempts in this area are discussed, with thoughts on how to develop this into a more general methodology.