The Future of National Distributed Control Systems

E-mail Wayne Manges

You can't define networked embedded control of cyber physical systems without factoring in the concept of fragility.

The NITRD Workshop
What will distributed control systems look like in twenty years? The National Coordination Office for Networking and Information Technology Research and Development (NITRD) held its workshop—led by UC Berkeley's Shankar Sastry, with help from Helen Gill of the National Science Foundation and representatives of several other agencies—in Pittsburgh on November 8-9 to begin to answer that question. The NITRD has been looking at "high confidence" software and systems for the U.S.'s post 9/11 critical infrastructure. NITRD is composed of a group of government agencies that includes the Department of Energy, National Institute of Standards and Technology, National Science Foundation, and National Security Agency. The goal of the workshop was to help these agencies create a suitable research agenda for future networked embedded control for cyber physical systems. My goal was to show that wireless systems could play a role in such high-confidence systems.

High Confidence Through Metrics
Throughout the proceedings, I emphasized the need for analytical performance metrics and critical definitions. To do this, I started with the concept of high confidence.

Since the incident at Three Mile Island, when an operator took control because he did not trust the automatic systems, many recognize the importance of operator confidence. As I thought about how wireless will play in these high-confidence systems, I concluded, once again, that we need to include real-time performance metrics that allow an operator to determine the status of the sensing and control networks with sufficient analytical rigor to ensure the necessary confidence.

The term I introduced for this in an earlier Sensors Wireless Networking Newsletter was fragility. Fragility is a measure of confidence in a system that is a function of the difference between expected performance and actual performance. Expected performance can be based on analytical models or past performance.

The solicitation that will be coming out of this workshop will undoubtedly include something about analytical measures associated with performance. The idea of using commercial networks to control distributed resources still sounds scary to me, but my guess is that it's coming.