Model-based Connguration of Machine Control Software Extended Abstract Model-based Computing Model-based Software Connguration

Following the trend of mass customization, repro-graphic machines (photocopiers, printers, fax machines, etc.) are increasingly designed and assembled from standard components. For example, a system engineer may design a paper path for a new print engine conngu-ration from components such as sheet transports, sheet inverters, gates, registration units, and trays. Similarly, customers may choose from and plug together diierent complex modules such as mark engines and feeder and nisher modules in order to get the desired function-ality. One of the most complex steps in this composition of reprographic modules is the automatic conngu-ration of the system control software. In this extended abstract, we give an overview of our approach to this task. See (Fromherz, Saraswat, & Bobrow 1999) for a complete review. When planning and scheduling the machine's operations (e.g., feeding, moving, printing, inverting sheets), the system control software has to satisfy each mod-ule's physical and computational constraints (timing, sheet sizes, etc.), and take into account the often complex interactions of the modules' behaviors. For this task, we have deployed an approach to developing real-time system-level controllers for electro-mechanical machines that consists of three concurrent activities: (1) the development of application-independent, declarative , constraint-based models of physical machine modules and conngurations, (2) the development or re-use of a separate, connguration-independent control software architecture, and (3) the development or re-use of mediating reasoners that provide the glue for embedding the models into the control architecture. In this model-based computing approach, system models become an integral and executable part of the system software, enabling the software to adapt itself to diier-ent conngurations, and to exibly react to changes in the system's capabilities. Model-based computing relies on the use of domain-speciic constraints for modeling, an idea that is very familiar to engineers working in a particular domain. In fact, developing a suitable modeling language for this domain was crucial to getting our approach adopted by software engineers. We learned that engineers feel comfortable with a language that provides the domain-speciic constructs important for modeling, at the right level of abstraction, with a minimum of ballast needed for reasoning about the models, and with a familiar look and feel. The resulting Component Description Language , CDL, is a high-level, engineering-oriented language with domain-speciic constraint systems. CDL models describe a device's capabilities through in-put/output transformations and constraints on moving parts, timing, and resources. However, it can be diicult to provide a simple semantic interpretation to …