Significant pieces of our nation's critical infrastructures depend on automation systems that operate under tight safety and timing constraints. Critical infrastructure security, bottling and packaging, material and container handling, automated manufacturing, on-board ship systems, locomotion systems, airport baggage handling, and amusement park rides are a few examples of such systems. With over a trillion dollars of installed base in North America, and over $90 Billion dollars forecasted revenues for year 2008, these systems represent an important class of embedded real-time systems. However, there are serious design and operational problems in existing automation systems. Rigid architectures and proprietary/inflexible implementations have resulted in systems that are difficult to operate and maintain. This project addresses research issues that are critical for incorporating advances in networked embedded computing into automation systems with flexible topologies. Technology support for atomic and coordinated actions enables cooperative operation in complex, coupled automation systems. A real-time middleware framework for sensor-actuator systems is being developed, providing services for node-level execution, scheduling, synchronization, mode-management, and fault management. By incorporating these advances, it is possible to reduce costs, achieve fine-grained control, improve safety, and design automation systems with flexible topologies for a variety of applications such as automotive assembly lines, chemical process industries, warehouses, airport baggage carousels, amusement park rides and package distribution centers. By focusing on the specific domain of automation systems, this project seeks to drive several innovations that include: targeted integration of sensors and actuators in low-cost nodes, dynamic re-configurability of wireless sensor networks driven by exception conditions, exploitation of redundancy to mask individual node or sensor failures, a scalable infrastructure, monitoring of large-scale operations, performance under constrained conditions, specialized algorithms and communication protocols that address real-life automation requirements, and inter-operability with current automation systems.
Wireless Sensor Networks (WSNs) represent a relatively new frontier of technology development. This domain has enabled small microcontrollers with limited computing power and wireless communications capabilities but operating off of battery power to obtain sensory data about the operating environment (obtaining information like temperature, humidity, sound, light, vibration) and communicating them over longer distances by relaying this information from node to node. This information can then also be used to make control decisions like turning off lights in an empty room, raising an alert that a window is open during a wintry night, excessive power is being consumed by a leaky refrigerator, and a laserprinter not going into a low-power sleep mode due to the toner running out. In this research project, we have developed multiple generations of a wireless sensor network platform (or node) called FireFly, and a mini real-time operating system called Nano-RK that runs on these nodes. Information about the hardware and software platform can be found at www.nano-rk.org. These platforms have been used by many other researchers both in the US and abroad, enabling rapid proliferation of such technologies for additional development and deployment. We have also used these nodes to create a large-scale network called Sensor Andrew, that has instrumented 5 different buildings on the campus of Carnegie Mellon University. This deployment has enabled many new capabilities: determine the position of (cooperating) users within these building which lead to many kinds of "location-aware" services ,monitor and control the energy consumption of various rooms, diagnose online any problems with the system, and to archive and view historical information such as temperature, light and vibrations in these buildings. We have also evaluated industrial manufacturing testbeds by sensing and controlling assembly, machining and conveyor operations. Lastly, we have also developed and deployed mechanisms to actuate (turn on/off) appliances remotely, and to time-synchronize various nodes across a building using the phasing information available everywhere where 110/220V electricity signals are present.
