Model-based computing: Developing flexible machine control software

In the conventional approach to simulating, controlling, and diagnosing a r eal-world physical system, engineers typically analyze the interactions of the system's components and processes, and then develop new and dedicated c ode for that system. Instead, building on principles from model-based reaso ning and constraint programming research, we propose an integrated approach to software development we call model-based computing. We present this app roach in the context of control software for modular electro-mechanical sys tems. Our approach is used in commercial systems and has been shown to both simplify the development of machine control software, and make the softwar e and the controlled systems more flexible and effective. In this paper, building an a generic control software architecture, we firs t develop a domain theory with corresponding modeling language. Models capt ure a system's capabilities from first principles and independently of the control task. We then introduce modeling technology using concurrent constr aint programming, which gives our modeling approach a sound and powerful th eoretical foundation. Constraint programming also brings with it a host of generic reasoning techniques such as deduction, abduction, and search, and we show how such techniques can be applied to the model-based configuration and control of our systems. We end with a review of how model-based comput ing can be extended to other tasks such as design and testing. We believe t hat together, models, task architecture, and reasoners offer a compelling f ramework for building software for computationally controlled systems. (C) 1999 Elsevier Science B.V. All rights reserved.