Main menu >> Computer Science >> Real-Time Systems
Computer Science
Real-Time Systems
- Customizable Coordination Services for Large-Scale Network Embedded Systems
- Defect-Tolerant System Integration and Evolution
- Real-Time Infrastructure for Context-Dependent Networked Agents
^ Customizable Coordination Services for Large-Scale Network Embedded Systems
G. A. Agha,* P. Chang, P. Thati, R. Ziaei
Defense Advanced Research Projects Agency, F49620-97-1-0382
The focus of this research is on developing application independent services to coordinate large network embedded systems. The coordination services will use customization and composition to enable dynamic adaptation in uncertain environments. The approach is to define algorithms that are based on stochastic models of system behavior, which enable the research team to represent the incompleteness in information about the current global system state as well as the unpredictability of the environment. The operational model uses probabilistic transitions rather than simple nondeterministic interleavings of actions, and it explicitly accounts for duration of transitions. The goal is to develop algorithms that provide for coordination in real-time and that guarantee the desired properties with sufficiently high probabilities. Examples include algorithms for approximate consensus (such as approximate synchrony), recovery, and hierarchical coordination. The algorithms will be implemented to provide a code basis for application independent coordination services. The implementation strategy is to build a repository of basic coordination services using reflective middleware. The goal will be to derive more complex algorithms based on simpler core resource management services.
^ Defect-Tolerant System Integration and Evolution
L. Sha,* J. Lopes, C. Parrott, K. Wehner
U.S. Office of Naval Research, Sha 2063
Large software systems are developed by integrating software components. Unfortunately, many complex software components often contain defects. On the other hand, the technology exists to develop modest size software components with a high degree of confidence. Flight control software is an example. This research focuses on algorithms and architectures that can leverage simple high assurance components to ensure the integrity of large distributed real-time systems, in spite of faults in complex software components, to support reliable upgrade of and online tuning of deployed systems and to detect and resist cyber attacks.
^ Real-Time Infrastructure for Context-Dependent Networked Agents
L. Sha,* X. He, X. Liu, M. Tian
Electric Power Research Institute
The control of the next generation of power generation and distribution systems will utilize networked control agents whose actions will be context dependent. This work will provide a distributed real-time computing infrastructure for agent-based control. The research focuses on ways to provide real-time, coherent views to distributed control agents; support the fault tolerant, dynamic replacement of control agents; and check to determine if an agent is compromised by a computer virus.
