Real-Time Sensor-Actuator Networks

Emerging technologies offer new paradigms for computation, control, collaboration, and communication. To realize the full potential of these technologies in industry, defense, and homeland security applications, it is necessary to exploit the real-time distributed computing capabilities of sensor-actuator networks. To reliably design and develop such networks, it is necessary to develop deeper insight into the underlying model for real-time computation and the infrastructure at the node level that supports this model. In this paper, we discuss a new node-level operating system and mechanisms necessary to deploy reliable applications. The overriding issue that guides the design of this operating system is quality of service metric called predictability. A sensor-actuator network is a distributed platform for integrated computation and control in real-time environments. The nodes in such a network are distinguished by being resource constrained. The power of the network arises from the interactions between simple nodes. Such a network extends the popular distributed sensor networks in several dimensions. After identifying a real-time model, we develop a notion of predictability for a sensor-actuator network. We discuss how the node-level operating system is designed in the resource-constrained environment. An efficient multithreading mechanism and scheduling strategy are required to ensure that local tasks are executed within jitter bounds and that end-to-end delays do not violate application constraints. Mechanisms to support communication, monitoring, safety, fault tolerance, programming, diagnosability, reconfiguration, composability, interoperability, and security are discussed.