Cyber physical systems: Connecting computation and the physical world

A perspective from Keith Marzullo, Division Director for Computer and Networked Systems at the National Science Foundation

CPS Image

Image Credit: Nicolle Rager Fuller, NSF

Imagine a world in which our physical environment interacts seamlessly and intelligently with us:

  • Our homes and offices respond to our needs while conserving our use of resources such as energy and water;
  • We have access to autonomous vehicles that provide transportation while reducing congestion, fuel costs, and accidents: and
  • We have medical implants personalized for our bodies and conditions that can communicate with our medical providers in a way that improves our health while maintaining our privacy.

Until recently, this kind of world existed only in the realm of science fiction, but no longer: prototypes of the systems described above are being built in research labs around the world. We call such systems cyber-physical systems, or CPS for short. In 2008, the National Science Foundation (NSF) created a x program to develop the engineering knowledge needed to build such systems, along with the necessary underlying scientific and mathematical techniques and the platforms that support their development and deployment.

Cyber-physical systems are "smart" networked systems that are engineered to sense and interact with the physical world, and to support real-time performance in safety-critical systems. With CPS, the joint behavior of the "cyber" and "physical" elements of the system is critical. Computing, control, sensing and networking are deeply integrated throughout the system, and the actions of components and systems must be carefully orchestrated.

Cyber-physical systems interact with people as well, both as users controlling them and as part of the environment in which the systems operate. Because these systems control aspects of the physical world, they are life-critical. Such systems must be designed to be correct, which requires formal models, verification methods and the ability to be certified.

The need for CPS arises in myriad situations. For example:

  • Transportation: Many of the long-haul passenger flights we take now are close to unmanned – there are pilots, but they may not need to do anything outside of being available if needed. As we continue to make progress in unmanned flight, some see it being used, for example, for short taxi-like trips. Semi-autonomous and autonomous automobiles are also being developed to reduce automobile crashes. More than 90% of the approximately 6 million automobile crashes in the United States are caused by human error, and 1 in 5 crashes are due to distracted drivers. Semi-autonomous automobiles are predicted to be commercially available in a few years, but to get to fully autonomous automobiles that can operate with human-controlled cars, pedestrians and so on requires significant research in CPS areas.
  • Manufacturing: The complexity of what we manufacture in the United States and the products we are able to design and manufacture are constantly increasing. The modern assembly line is itself a cyber-physical system, with automated machines and humans working together to deliver products to a supply chain controlled by computer algorithms. Increasingly, the products that are being produced in factories – including systems for the military – are also CPS. For this reason, understanding CPS at a fundamental level is vital for designing and manufacturing smart and secure systems.
  • Health: As in many countries, the population in the United States is getting older. Aging, coupled with the rise of chronic, lifelong diseases like cancer, diabetes, and arthritis, is giving rise to medical and health care issues. By combining inexpensive sensing, ubiquitous communication and powerful computation, cyber-physical medical devices and health monitoring and maintenance products stand to revolutionize care and wellness.
  • Sustainability: With a projected global population surpassing 9 billion people by 2050, an uncertain and changing climate, and up to 50 percent of food lost between production and consumption, our agricultural systems need to be more autonomous and efficient. CPS technologies supporting both conventional and vertical farming (urban farming in skyscraper greenhouses) and food logistics (moving food from producers to consumers) are key to increasing efficiency throughout the value chain, improving our environmental footprint and securing our food supply.

There are many other examples of situations that benefit from CPS, including: energy, building control, military and national defense, emergency response, global commerce logistics and "smart cities" that encompass many of these situations. The potential macro-economic benefit of the development and deployment of CPS in the coming decades is enormous.

The motivations for CPS come from different sectors, but many of the technologies face a common set of fundamental challenges in the science and engineering of complex systems that conjoin the cyber and the physical. By abstracting from the particulars of specific systems and application domains, the NSF CPS program seeks to reveal cross-cutting fundamental scientific and engineering principles that underpin the integration of cyber and physical elements across all application sectors. Since 2008, NSF has invested more than $250 million to conduct the basic research that underlies all CPS systems.

A great example of research supported by the NSF CPS program is the autonomous vehicle – a Cadillac SUV – developed by Carnegie Mellon University (CMU) researchers over many years. This past June, 18 lawmakers took short rides in this self-driving Cadillac around the Capitol in Washington, D.C. and to the Pentagon and back.

The passengers were initially nervous, but they adapted quickly to the novelty of the experience and enjoyed the trip. Though it may look like any other SUV on the road, the research that went into the vehicle includes breakthroughs in sensing (traffic, obstacles, pedestrians, cyclists, etc.), control (for obstacle avoidance, for merging, etc.), and verification of the system. Each of these fundamental research advances contribute to a CPS engineering discipline. The projects that NSF has funded, such as the CMU autonomous automobile, are impressive. Yet despite the progress we have made, we are still a long way from having a mature systems engineering for high-confidence CPS. The state-of-the-art design for embedded systems is not sufficient for large, distributed applications that need to accommodate complex environments. CPS also requires advances in managing autonomy and cooperation – both with humans and with other CPS – in order to meet high standards for safety, security, scalability, and reliability. We simply don’t know yet how to build verifiable systems that can interoperate at the scale of future CPS.

Of course, NSF isn’t the only agency interested in CPS. Federal agencies coordinate CPS research in part through the Networking and Information Technology Research and Development (NITRD) Program. As part of this coordination, in 2014, NSF partnered with the U.S. Department of Homeland (DHS) Security Science and Technology Directorate (S&T) and the U.S. Department of Transportation (DOT) Federal Highway Administration (FHWA) to announce a joint funding opportunity that identifies basic research needs in CPS common across multiple application domains, along with opportunities for accelerated transition to practice. Moreover, advancing CPS requires partnerships spanning the Federal government, academia, industry and beyond.

It has been an amazing six years for the NSF CPS program. Personally, I have enjoyed riding in the CMU car (being driven in reverse at highway speed was perhaps more shocking than enjoyable), watching a copter gracefully swoop down to catch a tossed ping-pong ball in a bowl on its top (in spite the turbulence being kicked up by the copter) and learning about the complexities of designing cardiac pacemakers for individual patients. But these are just the beginnings of what CPS will offer.

Programmatically, we have made great strides in many areas: control, verification, timing, scalability, robustness, human interaction, dealing with uncertainty, and many others. The CPS research community shows tremendous initiative and innovative abilities. While there is still a long ways to go, this community is creating the world that we could only imagine a few years ago.


Stay tuned for the results of this cross-agency effort. To learn about developments in NSF’s CPS efforts, visit the Cyber-Physical Systems Special Report and follow us on Twitter .