This Information Technology Research
(ITR) project is aimed at developing the foundations of a modern systems science that is simultaneously computational and physical. Traditional systems science has applied mathematical modeling to the design and analysis of physical systems, such as circuits and mechanical systems that are designed to perform some feedback control or signal processing function. Such systems are found in automobiles, aircraft, communications devices such as radios and telephones, toys, and weapons, for example. But increasingly, such systems are realized using embedded software, and the traditional systems science poorly models the realities of that software. Conversely, the science of computation, which addresses software, is traditionally about procedures that transform data. It has difficulty with software that engages the physical world. The abstractions that are used minimize the role of time and concurrency, both of which are critical features of the physical world. This project will match mathematical modeling and design to embedded software.
The modern systems science we aim to develop represents a major departure from the current, separated structure of computer science (CS), computer engineering (CE), and electrical engineering (EE): it reintegrates information and physical sciences. It will have a profound impact on teaching and research, and is committed to re-architecting and retooling undergraduate education.
This project seeks to provide new engineering methods for the design of complex embedded systems that are reliable and robust to partial system failures, and yet leverage the cost reductions and improvements in sophistication made possible by software technology. Without such reliability and robustness, the potential of software cannot be realized in safety-critical applications, such as medical devices, automobiles, and aircraft.
The project has four focus areas of research:
- Hybrid systems theory.
- The focus here is on scaling up
pioneering approaches that integrate physical modeling with
computational systems. Existing methods work for simple,
low-dimensional systems; the researchers will seek methods that apply
to complex, interconnected systems with stochastic attributes. As part
of this effort, the mathematical foundations of systems theory need to
be rebuilt in a way that tightly integrates continuous and discrete
domains.
- Model-based design.
- The main effort here is to develop a set of models with solid
mathematical foundations that allow for the systematic integration of
diverse efforts in system specification, design, synthesis, analysis
and validation, execution, and design evolution. Key to this effort is
research in concurrent and real-time models of computation, and in
formal ways of modeling the modeling techniques themselves.
- Advanced tool architectures.
- The deliverables from this project will be a set of reusable,
inter-operating software modules, freely distributed as open-source
software. These modules will be toolkits and frameworks that support
the design of embedded systems, provide infrastructure for
domain-specific tools, and provide model-based code generators.
- Experimental research.
- The program will leverage existing system-building efforts
involving avionics, anti-terrorism technologies, vehicle electronics,
and autonomous robots. In addition the project will apply its methods
to networks of embedded systems for applications such as environment
monitoring, building protection, and emergency response.
The lead partners are the Center for Hybrid and Embedded Software Systems (CHESS) at the University of California at Berkeley (UCB), the Institute for Software Integrated Systems (ISIS) at Vanderbilt (VU), and the Department of Mathematical Sciences at the University of Memphis (UM).
