Font Size: a A A

Distributed control for embedded system design

Posted on:2009-11-25Degree:D.EngType:Dissertation
University:University of California, Santa BarbaraCandidate:Hoover, Gregory WFull Text:PDF
GTID:1448390005454709Subject:Engineering
Abstract/Summary:
Embedded systems are a complex tapestry of hardware and software components sourced from numerous and varying outlets. Modern systems push the limits of system integration, combining heterogeneous components from diverse fields ranging from bio-chemical polymers to microscope mechanic systems. Meeting the constraints of these systems requires novel architectures that offer greater functionality at higher performance and lower power. At the same time, methodologies are required to bridge high-level system behaviors with the implementation platform, providing improved hardware support for software execution and practical communication strategies. An essentiality of realizing these systems are component-based design flows that allow composition of intellectual property (IP) while helping the designer cope with expanding control and communication complexity. To date, many techniques reduce system behaviors to monolithic control architectures that unnecessarily synchronize tasks, resulting in suboptimal design implementations.;An alternative, distributed, control describes behavioral tasks independently according to a control and data communication model that allow mapping to a variety of physical resources. Supporting this model are latency-insensitive interfaces that allow composition of behavioral tasks with guaranteed system correctness. Designer effort is reduced by extending the latency-insensitive model up to the specification level, where a rule-based language allows functional design reasoning without the overhead of specific temporal constraints. A semi-automated design flow enables design realization in a number of technologies in both hardware and software. These designs potentially exhibit fewer synchronization points than their conventional counterparts, improving performance and power efficiency, while simultaneously increasing execution parallelism. Added parallelism motivates a practical need for exploration of architectures and added physical resources.
Keywords/Search Tags:System
Related items