Unified ardware and Software MO mart Syste esig

“Smart” electromechanical systems (e.g., washing machines, car anti-skid braking systems and reprographic machines, i.e., printers and photocopiers) are machines with intelligent, embedded computer control. Conventional embedded controllers performed a preprogrammed set of actions, but as machines have become more versatile, controllers are being designed to generate and evaluate different plans to perform specified tasks, search for optimal actions, and react to changing environments. Traditionally, hardware design in embedded systems has preceded software design resulting in sub-optimal designs. As the use of embedded software proliferates, integrated hardware and software design becomes critical. Like conventional design, smart system design explores the space of possible solutions to develop one that satisfies functional and performance specifications and that optimizes desired criteria such as speed, cost, and power consumption. Smart system design is a difficult problem because decisions must be made about allocating functionality between hardware and software such that the combined design performs correctly and optimizes desired criteria. Furthermore, anytime algorithms must be designed where both planning and execution time affect overall performance. Finally, because there are dependencies between the hardware and the software, the exploration of the design space must consider both the hardware and the software. Adopting an integrated approach over the present, sequential methodology, will improve quality and reduce the time taken to design complex, smart systems. A key to exploring the design space effectively, is to develop a unified representation of hardware and software at suitable levels of abstraction. [l] has developed discrete hardware models for scheduling sheets in reprographic machines. We propose a unified, hybrid modeling scheme that integrates discrete hardware and