U.S., China Collaborations in Computing and Sustainability

October 3rd, 2011 by Erwin Gianchandani Post a comment »

Fred Roberts, Director, Center for Discrete Mathematics and Theoretical Computer Science (DIMACS), Rutgers [image courtesy Rutgers].Stephen Greenfield, Rutgers [image courtesy Rutgers]This is a special contribution to this blog by Fred Roberts, director of the Center for Discrete Mathematics and Theoretical Computer Science (DIMACS), and Stephen Greenfield, Professor of Mathematics and a member of the Graduate Faculty, at Rutgers. The pair organized a workshop entitled “U.S.-China Collaborations in Computer Science and Sustainability,” bringing together 45 U.S. and Chinese computer scientists, mathematicians, ecologists, and representatives from other disciplines for a two-day meeting Sept. 19-20 in Piscataway, NJ. The full program, including list of participants and talks, can be found here. In this post, Fred and Stephen summarize the workshop, including key findings.

The Sept. 19-20 workshop at Rutgers followed a series of U.S.-China Computer Science Leadership Summits featuring leaders of major U.S. and China computer science departments and research centers. Those Summits, initiated at the suggestion of former U.S. National Science Foundation (NSF) Assistant Director for Computer and Information Science and Engineering (CISE) Peter Freeman in 2006, discussed major areas of research and education of common interest to both countries. At the third Summit in Beijing, the agenda moved from rather general topics to an exploration of potential research collaborations, through an emphasis on topics that are of concern to both societies: health care, environmental challenges, security, energy, economic development. At that third Summit, it was agreed that “next steps” should include identification of truly collaborative projects — and, in particular, computer science and sustainability was selected as one topic ripe for initial collaborative projects.

The growing human population and increasing pressures for development have led to a variety of challenges for life on our planet, resulting in the fundamental question of whether current patterns of human activity are sustainable. Human activity is closely tied to the natural environment and there is a two-way interconnection between human activity and environmental processes. Increasingly, we are noting how human activities affect the systems that sustain life, including climate, healthy air and water, availability of food, etc. Indeed, the earth has finite resources that humans require to sustain current lifestyles: sources of energy, clean water, arable land, etc. As environmental conditions change, there are possibilities for new diseases, species can migrate into non-native areas, crowding out the species to which we are accustomed, and the human condition can be threatened by environmental change. Fundamental societal structures such as national boundaries and the health of our economic systems can be affected by competition for changing natural resources, including shortages and, in some cases, surpluses. These problems are complex, multi-disciplinary, and intertwined — and, most importantly, they do not respect international borders. Herein lies the motivation for the call for a “science of sustainability” primed for U.S.-China collaboration.

Recognizing the challenges to life on our planet as we know it, the NSF has initiated a major new initiative on sustainability (officially called Science, Engineering, and Education for Sustainability, or SEES) that involves all the directorates in the Foundation. In particular, computer science has a major role to play in helping to address the challenge to sustainability. Moreover, NSF’s major international research program, Partnerships for International Research and Education (PIRE), will emphasize SEES in FY 2012. The DIMACS workshop sought to describe these computer science challenges and to emphasize opportunities (including topics) suitable for U.S.-China collaborative projects, both research and educational. The problems of sustainability cross many disciplines. The computer science challenges the workshop identified can be expected to lead to research on a wide variety of topics of great societal importance such as climate change, environmental health, management of limited natural resources, and the interconnections of these topics with healthy economic systems and enduring social structures. This workshop follows a series of workshops held in the U.S. on computer science, mathematics, and sustainability:

  • Toward a Science of Sustainability (November 2009) (see the report here);
  • Mathematical Challenges for Sustainability (November 2010) (see the report here);
  • The Role of Information Science and Engineering in Sustainability (February 2011) (see the report here); and
  • IEEE SECON Workshop on Information and Communication Technologies for Sustainability (report forthcoming).

The keynotes

At the U.S.-China workshop, three keynote lectures were given describing central ideas allowing computer science to support and further the new discipline of sustainability. Zhiwei Xu, Deputy Director of the Institute of Computing Technology of the Chinese Academy of Science, talked about the “Human-Cyber-Physical University.” He pointed to three drivers of CS and sustainability: industrialization (and the possibility of “smart equipment”); urbanization (and IT help for a sustainable life style); and informatization (moving from CS in cyberspace to CS at the interface between cyberspace and physical space — a “ternary computing” of the future). CCC Council member Randy Bryant, Dean of the School of Computer Science at Carnegie Mellon University, described “Computer Science Research Opportunities in Sustainability,” including energy, transportation, environment, and climate. Bryant discussed the relationship of future systems that continuously monitor themselves, adapt, and repair themselves; systems designed as networks of loosely coupled agents; CS serving the needs and characteristics of people; and trustworthy modeling and simulation. Xiaoming Li, Assistant President and Director of the Institute of Network Computing and Information Systems at Peking University, gave a survey of sustainability activities at his university, including projects on water (flood control, river ecology rehab); species preservation; grassland ecosystems; solar energy; and conservation biology. He said that computer science has an important role to play, but not many Chinese computer scientists are involved in sustainability yet.

Additional lectures discussed topics such as smart/green buildings, green IT, alternative energy, software, sensors, the environment, and climate change. Additionally, five discussion groups met and reported on opportunities for further work, including multinational collaboration. A representative of the NSF was present to describe the Foundation’s strong interest in fostering multinational collaboration on questions of sustainability, such as through PIRE. We were told that both the U.S. NSF and the National Science Foundation of China (NSFC) have great interest in supporting research in sustainability.

Some general themes were apparent throughout the workshop:

Massive data sets: collection and analysis

Almost all of the current and planned research related to sustainability involves the collection and analysis of truly massive data sets. This is what Bryant calls “Data-Intensive Super Computing.” Since disks that store a terabyte can now be purchased for less than $100, storage itself is no longer an obstacle. But the analysis of such data in real time cannot use many classical methods: too much space and too much computation would be necessary. Instead, new algorithms are needed.

The ability to collect massive amounts of data about the state of our environment presents both a wonderful opportunity and a great challenge for sustainability science. Even implementation of realistic data collection will need new methods. We can now “create” easily a local area network with 10 to 20 computers and other devices. As they “wake up,” they recognize each other and organize themselves. But the scale of data collection involving sustainability resembles such an activity magnified by a large factor. For example, collection of information about temperature and pollution levels in a large building or over a significant area could involve tens of thousands (or hundreds of thousands!) of sensors. How should this “network” be organized? How can it be constructed and deployed economically and reliably?

Questions concerning such data accumulation include privacy and security issues. Here are a few simple examples, some almost ludicrous: Should everyone know the water usage in each family unit? Should the number of toilet flushes be part of a public data set? What sort of information should be collected and how private should it be? Should “everyone” know how you drive to work or what public transportation you characteristically use? It can be argued that aggregation of such data would allow better resource allocation, but how much privacy should be preserved? Consider electric generation and distribution networks (the “grid”). Certainly sustainability questions will be intimately concerned with the reliability and redundancy of the future networks known as “smart grid.” But knowledge of the weaknesses and costs of such a network could compromise the security of a city, a region, or a nation. And the ability to incorporate “smart meters” could allow outsiders to learn, from our pattern of energy use, what movies we are watching.

The need for interdisciplinary work

Computer science will be successful in sustainability investigations only as much as the computer scientists involved are willing to work with and listen to experts from other disciplines. If we wish to study and foster species preservation then mammals, birds, reptiles, fish, and even insects will all need detailed and distinct study: what “territories” (in space, time, resources, and other species) are necessary for threatened species to be preserved, not tenuously, but with a good chance that they will exist in the far future? Different but allied questions need to be considered in each case. A “theoretical” analysis might be interesting, but intervention by experts will be needed for meaningful results. Similar work with material scientists and engineers is necessary to make progress with green/smart buildings, buildings that use less energy and have safer interior environments than classical structures. Every project described at the meeting with useful consequences involved interdisciplinary research with engineers, scientists, and  social scientists (e.g., the cost and economics of health regulation).

The implications for education and professional preparation are immediate. At the K-12 level, delivery of information at appropriate levels needs to be developed. In college and in graduate and postdoctoral study, individuals who wish to use computer science to study the problems of sustainability should be encouraged if not required to obtain expertise in an allied field. But how do we fit in such interdisciplinary education when there is already too much to learn about one’s home discipline?

New algorithms, or, better, new types of algorithms

In addition to the scale of data acquisition and analysis, as already mentioned, another characteristic that emerged repeatedly in reports of projects and suggestions for study is that the algorithms needed will have to work with data sets that are subject to uncertainty or error — so the collection, analysis, and even conclusions will not be of the form “do this” but instead will be stochastic: “the likely desirable outcome will occur with the following distribution of probabilities if these suggestions are implemented.” Already the beginnings of a class of algorithms dealing with this situation are appearing, a mixture of computer science, statistics, operations research and mathematics. But this is only a beginning.

Also, such examples as climate change as well as species preservation show that the analyses will need to reflect “non-stationarity.” The initial characteristics of probability distributions of the data, which may be derived from existing data sets or from modeling assumptions, will almost surely change as time (and other variables) change.

Modeling and simulation on the scales needed is difficult and again will need new ideas. Experimentation in many cases is just impossible (“build two power plants here instead of a single large plant there and see what happens”).

Informing policy decisions

Communicating the needed hypotheses and caveats will be difficult. The history of climate change discussions, as well as those regarding such goals as species preservation and pollution reduction, support this assertion.

Many of the questions involving sustainability will likely ultimately involve decisions by policy makers having minimal technical backgrounds, and communication will be difficult. Again, the controversies involving climate change, including analysis of the data and suggestions made from this complex analysis, are one example. Progress must be made in correct communication of complex ideas and analyses. We need to prepare computer scientists to communicate effectively about conclusions and recommendations, and the limits of these.

After the discussion groups reported near the close of the workshop, desires for  meaningful collaboration between groups in the U.S. and China were strongly expressed. One Chinese participant stated, “I didn’t fly thousands of miles just to give a half hour talk,” and the participant wanted to plan immediately for submission of parallel requests for support of such activities from the U.S. NSF and NSFC. In the end, several concrete ideas for research and educational collaborations resulted and many contacts for the beginning of networking were made.

Editor’s note: What do you think about the key points from the workshop described above? Share your thoughts in the comment space below!