In late 2006 CenSCIR researchers began to design and implement a campus-wide testbed for new sensing technologies and applications. The project's goal will be to make Carnegie Mellon one of the most "sensed" campuses in the world. By late-2007, Carnegie Mellon researchers will have access to a "living testbed" of sensors and sensed data to provide a platform for future research.
The U.S. infrastructure is a trillion dollar investment, defined broadly to include road systems and bridges, water distribution systems, water treatment plants, power distribution systems, telecommunication network systems, commercial and industrial facilities, etc. In spite of the enormous investments made in these systems and their importance to the US economy, we (government, industry and academia) are not very good stewards of this infrastructure. By their nature, infrastructure systems are large-scale, networked systems with physical components that may be themselves networks of systems and whose health, due to use, environment, and abuse, can significantly deteriorate. Because of expense and growing local demand, these systems have expanded over decades in a more or less ad-hoc fashion. Because of their size and highly interconnected nature, the operating conditions of the overall network are difficult to assess from local data. Often, local actions may give rise to unexpected global behavior.
Sensing technologies can be deployed to improve the performance and/or reduce the life-cycle cost and societal impacts of all life-cycle phases and over a broad range of physical infrastructure systems. The life-cycle of a constructed facility affords many opportunities for sensor-based decision support systems, from the first stages of construction through the demolition and reclamation of the construction materials. We must take better advantage of sensors and other types of technology to improve our construction, management and operation of infrastructure systems.
The Carnegie Mellon Center for Sensed Critical Infrastucture Research (CenSCIR) is a problem-driven center researching the many systems-oriented issues related to delivering actionable information about the condition and usage of our critical infrastructure systems. There are a number of researchers in the center working on research projects related to detecting defects during building construction, monitoring localized structural damage such as cracks in steel members, detecting and classifying sewer pipe defects from the many hours of video images collected from these sewers, support for first responders, etc. To better support these various researchers in the center, CenSCIR is building a “living-laboratory” for exploring the many systems-oriented issues related to delivering advanced condition and usage information about our critical infrastructure systems This testbed will provide a variety of real-world infrastructure challenges to researchers. Experimentation in laboratory conditions can provide valuable research, but the heterogeneous environment and unpredictable conditions in which real-world infrastructure systems operate will require a “living-laboratory” approach.
Therefore, Carnegie Mellon is undertaking a project named “Sensor Andrew”, a sensor network that will enable the dense instrumentation of the whole of Carnegie Mellon’s campus as a living laboratory for real-world infrastructure challenges.
A number of research questions will need to be addressed to achieve this vision of providing a nervous system for our critical infrastructure, including:
- What information needs to be measured about a system, and where and when should it be measured?
- What types of sensor are needed, but do not exist in a form usable for infrastructure applications? How are they powered for long periods of time?
- How do they remain functional for the long lives of most infrastructure systems?
- How should this extremely large amount of information be represented, stored, managed and exchanged and who is responsible for the stewardship of this data?
- How does one predict global behavior, and more importantly the onset of an incident, from localized sensor information, and what inference algorithms are needed to infer the state of the system without having centralized knowledge of the network?
- What types of action need to, and can be taken to prevent abnormal behavior?
- How does one translate the mass of information collected from the distributed sensor network into intelligent decision support that actually helps the operators of these systems make the best possible decision under the circumstances?
- What are the economic conditions that make such an approach to delivering a “nervous system” for an infrastructure system economically viable?
- Does it make sense to use such a system only at certain times when problems are anticipated?
- What can we do at construction time and what can wait until later?
- Fully heterogeneous
- Different components
- Distinct data sources
- New merged data
- Different applications
- Non-centralized architecture
- Highly scalable and adaptable
- Not purpose-specific
- Easy to expand and reconfigure