Conclusions and Recommendations: An Agenda for Research
EmNets will be embedded everywhere, from automotive instrumentation to precision agriculture to battlefield surveillance. They raise fundamental research challenges in part because they will be performing critical functions and also because they are inherently distributed and tightly coupled to the physical world through sensors and actuators. Moreover, while they are rich in the numbers of elements, they are at the same time highly resource constrained in the capability of the individual elements. This chapter builds on the findings and discussions in Chapters 2 to 5 to specify particular research projects and processes that will be necessary to realize the vision articulated throughout this report.
As outlined in this report, EmNets present a number of research challenges that need to be addressed. An important message for the research enterprise is that new approaches to the study of systems rather than components must be developed as a deeper understanding of the emergent properties of many interconnected elements is gained. To attain this goal, research will need to become more interdisciplinary than ever before as practitioners learn to design, deploy, and—hopefully—trust these large-scale information systems. The need to approach the challenges presented by EmNets from a systems-oriented, interdisciplinary perspective stands out among the many technological problems delineated elsewhere in this report. Failure to meet this need would be the most serious impediment to realizing the full potential of EmNets in society.1,2
The growing complexity of information technology systems will be accentuated by the evolution of EmNets. This complexity arises not only from the large number of components involved but also from the lack of determinism and the continual evolution such systems will undergo. Effort on the part of the whole community (industry and academia, as well as funding agencies) is necessary. While there are specific EmNet applications emerging from industry, they do not encompass the kinds of scalable, robust, physically coupled EmNets that are discussed throughout this report. In the absence of appropriate funding, issues such as adaptive self-configuration, predictability, and computational models will not be addressed in ways that will enable comprehensive understanding. This lack of understanding will result in a technology that is both prohibitively expensive and prohibitively brittle and will preclude the widespread adoption of EmNets as envisioned here.
The Internet has provided one of the first real examples of a large-scale, heterogeneous networked system. It serves as an excellent model for observation and provides some early indicators of the issues arising from the widespread deployment of EmNets that will need to be addressed.3 The Internet consists of millions of loosely interconnected components that generate communications traffic independently of one another. There has been standardization in the middle levels of communication protocols, but a wide variety of physical interconnections, from optical broadband to wireless, is supported. However, from the casual user’s perspective, the degree of interoperability has essentially been limited to what can be done through a Web browser. For the most part, the currency of the Internet has been in the realm of information. The connections between today’s various information services are only now starting to evolve into multilayered and richly connected ensembles.4 Connections to the physical world have been limited to basic sensors (for example, cameras and weather sensors) and very few actuators (for example, camera motors and home remote control).
As noted throughout this report, EmNets will build on the Internet
experience (itself a product of significant federal research investment) but will also extend it in new directions. The physical world will be coupled to the information space. Sensors and actuators will be spread throughout the everyday environment. People’s activities will be recorded and affected by computing systems in virtually all spheres of life. The heterogeneity of the devices that will be interconnected will increase dramatically. From a world of PCs and servers, IT will move to smart dust,5 swallowable health monitors, and automated buildings. This move will require a much deeper understanding of how to build into EmNets the challenging properties of scalability and robustness.
In this chapter, several overarching research themes are described that draw on the discussions developed throughout the report. Following the description of these themes is a discussion of what will be required of the industrial and academic research enterprises in order to make progress on the substantive research recommendations made in this chapter and throughout the report. In addition, specific recommendations are made to federal funding agencies that, if followed, would facilitate progress in this area.
AN EMNET-SPECIFIC RESEARCH AGENDA
The committee has found eight key areas in which concerted research efforts are needed: predictability and manageability; adaptive self-configuration; monitoring and system health; computational models; network geometry; interoperability; the integration of technical, social, ethical, and public policy issues; and enabling technologies. This research will need to be very broad and very deep and so is unlikely to be achieved through industry efforts alone. Key to developing the research in these areas is the parallel pursuit of the major thrusts described in this report (see Chapters 2 to 5) and the integration of research across the various topics as necessary. Achieving progress in such a research agenda will require forward-thinking, visionary leadership and the willingness to invest in long-term research programs without requiring premature checkpoints or demonstrations and without a priori agreements on specific architecture, so as to allow room for reasonable exploration of the design space.
This section draws on the analysis contained in earlier chapters of the report to identify eight areas that should be part of such a research agenda.
These areas fall into three categories: (1) research that is needed to build robust and scalable EmNets, (2) research on social, ethical, and policy issues that result from the deployment of EmNets; and (3) research on component technologies that is unlikely to be addressed by the general IT research community.
It should be noted that networking is an implicit theme pervading most of these areas and so does not stand apart as a separate research issue. The success of networked systems of embedded computers will depend heavily on the networking research community and work going on there, including the work highlighted in Chapters 2 and 3. Progress in EmNets is not possible without progress in networking. The research issues raised by EmNets constitute a theme around which new networking research programs can be structured. Similarly, issues of usability and manageability arise throughout this discussion. The human element in complex, not-well-understood systems is critical at all levels, including design, programming, deployment, control, manipulation, and interaction. Human-centered approaches must therefore be incorporated into all of the research areas discussed below.
Predictability and Manageability: Methodologies and Mechanisms for Designing Predictable, Safe, Reliable, Manageable EmNets
Designing for predictability in EmNets requires new methodologies and design strategies that will support characterizable, understandable, and manageable systems. These systems need to allow for isolation of systems components and analysis of the interactions that take place within an EmNet that is exploiting massive amounts of interconnection. At the same time, methodologies are needed for presenting system behavior (including behavior that emerges throughout the lifetime of the system) to end users and system managers; these methodologies must transmit the correct information at the correct abstraction level. Users of EmNets may be experts at the task their computing system is helping them accomplish, but they should not need to know a lot about how the computing system is doing it. They need to be able to make certain basic inferences about what they can expect of their EmNet in order to make good, safe use of it.
It is likely that EmNets will radically alter the definition of a system. Instead of simply designing all the individual components of a system and their interactions specifically for a particular system function, people will be fielding components that provide basic capabilities. A “system” will mean exploiting the capabilities of those basic components in a new way by marshalling the capabilities of what is already deployed, altering
their function, or adding new elements. Pieces of a system deployed for one purpose may be utilized for other purposes not originally planned.
Moreover, continually changing or adding new elements to the mix will cause new, unintended behaviors to emerge. The Internet is providing some early examples of this: When new services are deployed, their increasing use may cause congestion and a decline in service quality at some points in the network. Once the network is embedded everywhere, every new deployment will probably trigger adjustments and possible detrimental effects on service only because it causes some contention for common scarce resources. Such behavior should occur in an understandable and reasonably predictable fashion. If something has broken, or even worse, is about to break,6 how should the EmNet inform its users?
EmNets must have interfaces that let users who are not professional system administrators wield them effectively, through normal as well as abnormal conditions such as partial system failures. Sets of abstractions should be developed that have meaning within the computing system itself yet still conform to users’ conceptions of the tasks they need to accomplish. EmNets have the same human computer interface problems as existing systems, exacerbated by the other, nontraditional aspects of EmNets, including users who are inexperienced with the intricacies of EmNets, real-time interactions with the physical world, long-lived systems that build user trust at the same time as their internal safety margins may be decreasing, and enormous overall system complexity.
Adaptive Self-configuration: Techniques to Allow Adaptive Self-configuration of EmNets to Respond to Volatile Environmental Conditions and System Resources in an Ongoing Dynamic Balance
EmNets will need to exhibit adaptive self-configuration in order to be viable. The massive numbers of elements, along with the resource constraints on individual elements and the environmental dynamics in which they will need to operate, combine to create a new and likely pervasive requirement for adaptive systemwide behavior that is unparalleled except perhaps in natural systems. The number of elements, resource constraints, and dynamics imply that systems cannot rely on a priori system design or manual adjustment. The system elements cannot simply be
configured to operate under worst case assumptions, because doing so would make them orders of magnitude less efficient and, in many cases, unable to meet system lifetime requirements. Moreover, EmNets cannot be dynamically configured centrally using global information because acquiring the global information consumes significant amounts of energy and is not scalable. Further, some of the adaptation will need to be done in a very short time frame, one that requires that processing of input and action be completed as quickly as possible to meet the real-time requirements of the application.
The current state of the art with respect to adaptation and configuration is exemplified in Internet protocols. These protocols are somewhat self-configuring and adaptive. However, they have not had to cope with intense input/output, environmental dynamics, and tight energy constraints as a primary design issue. EmNets will require the development of new distributed algorithms and techniques for provable distributed control. They will also require system models and characterizable behavior in order to support embedded systems with strict time constraints (latency, in particular). EmNets will need to provide rich interfaces to the application designers as well. For example, a truly scalable sensor network must self-configure so that the correct collection of nodes (those that have collected good signals from stimuli) collaborates in signal processing to detect and identify phenomena of interest inside the network. The particular sets of nodes that should participate cannot be determined a priori. Such a determination clearly depends not only on the nature of the application but also—and even more so—on the nature of the object(s) being monitored and the signals received by the nodes. EmNets will require nodes and their system interactions to be designed so that applications can influence the parameters and rules according to which nodes adaptively self-configure.
Monitoring and System Health: A Complete Conceptual Framework to Help Achieve Robust Operation Through Self-monitoring, Continuous Self-testing, and Reporting of System Health in the Face of Extreme Constraints on Nodes and Elements of the System
The mission-readiness requirements of EmNets will vary from one EmNet to another, but all will require a minimal amount of overall computational horsepower, a certain amount of interconnection bandwidth and latency, and some minimum amount of sensing and perhaps actuation. With current technology, this mission readiness will be evaluated by having the system perform periodic self-checks on all of those dimensions, with some kind of overall health indicated to the system user or administrator.
EmNets will change over time both in the numbers and kinds of their components and in the applications they are designed to perform. Current notions of system health, which tend to be based on the health of the individual components, do not extend to such systems, where no single component may be critical for the system to perform its intended function as long as the system can adapt to the current conditions. How such health, which is tied to the overall mission of the system rather than the function of the parts, can be defined and monitored by the system itself will be an important area of investigation. A critical challenge is that this system monitoring must be done in the face of resource constraints. For example, pulling system health information out of the system may consume valuable, unreplenishable energy. Just as the system may need to aggregate information about its function inside the network, it may need to aggregate information about its health.
Designing and constructing large systems of many heterogeneous components is already an extremely complex task. The added constraints of EmNets make it even more so. It may be possible to turn to fields such as economics, biology, and statistics for new tools to tackle this growing complexity.7 New approaches need to be developed for self-monitoring, self-testing, reconfiguration, and adaptation, as discussed in Chapters 3 and 4. Systems will have to be built with self-monitoring and self-regulating devices. Statistical approaches will be needed to properly detect situations requiring attention. Immune systems will need to be developed to counteract the unintended (or intended) effects of new deployments.
Because of the interactions with other requirements of the system, the conceptual framework for robust operation, adaptation, and self-testing cannot stand on its own. It must be part of a large conceptual model that takes into account the other features, requirements, and restrictions of the system, as discussed in Chapter 5. Research needs to be done not only on how to monitor and express this notion of system health, but also on the trade-offs that are possible between these requirements and the other requirements of the system.
Computational Models: New Abstractions and Computational Models for Designing, Analyzing, and Describing the Collective Behavior and Information Organization of Massive EmNets
Systems as complicated as EmNets will present enormous challenges for the analysis of behavior and performance. Existing tools and concepts
Various efforts to study complexity already reach out to a wide variety of disciplines. See, for example, the work of the Santa Fe Institute at <http://www.santafe.edu/>.
are barely adequate for understanding simple multiprocessor systems with four CPUs. They are clearly inadequate for systems with many thousands of physically coupled, long-lived, adaptable, self-configuring, interacting nodes. Moreover, defining the right model to handle these many components is not sufficient; the model needs to ensure that it is possible to reason about and understand the interactions of the various parts of the model so that appropriate trade-offs can be made, when necessary, in the design of the entire system.
In particular, in order to take better advantage of the many potential uses and impacts of EmNets, abstractions are needed for designing interactions with the physical world. Sensors and actuators will often play a key role in such systems. Moreover, new abstractions are needed for designing systems that make use of massive redundancy in order to deal with the extraneous data and uncertainty of the physical world. Unknown at this point is what building blocks will be used in EmNet environments that will play the seminal role that transactions and remote procedure call (RPC) played in more traditional systems. Defining appropriate data structures, process interactions, and APIs will require a substantial research effort, one that iterates between experimentation, concept development, and theory building.
The development of new abstractions for reasoning about collective behavior will be one of the biggest contributions of EmNets research (see Chapter 5). Both humans and the artifacts they design will require these abstractions to reason about and adapt to the new situations that will emerge when interesting new mixes of devices and services are created. Abstraction is one of the most powerful tools that mathematics and engineering have brought to the scientific enterprise. Each technological era has associated key abstractions. New eras bring new abstractions and vice versa. It is now time, as the era of EmNets commences, to begin the development of its principal abstractions.
Network Geometry: Ways to Support and Incorporate Network Geometry (As Opposed to Just Network Topology) into EmNets
In many traditional systems, the geographic location of a particular node is not important; instead, what matters is the abstract network topology. The fact that EmNets are coupled to the physical world requires understanding how to generate and use other forms of location information, such as three-space coordinates or logical coordinates associated with a building structure, for example. Such information can be both an important attribute of application-level data and a significant organizational principle for the system itself. When organizing information at the application level, knowing which nodes are in close physical proximity to
other nodes can be very helpful. For example, location information could be useful in determining coverage of a particular physical area. At the system level, such information can be used when trying to achieve efficient system behavior. For example, a node might be interested in determining the closest repository for storing long-term data. In such a case, close physical proximity is desirable in order to reduce resource expenditures. Location information is useful in another way as well: Using three-space information in combination with static environmental information allows the creation of logical location information that takes into account the surrounding environment.
As discussed in Chapters 2 and 5, global positioning system (GPS) technology is not sufficient for all of the network geometry needs of EmNets. GPS is a good model for the services needed in many outdoor, three-space-oriented systems but not necessarily for EmNets that are indoors, on the battlefield, or in other remote locations. Moreover, GPS is not ideal for networks whose nodes are small. New kinds of systems are needed that are not constrained in the way GPS systems are. Research into systems that can take into account the logical structure of the geographical environment—for example, walls separating offices, the location of doors, or the inside of a vehicle—is also essential.
Interoperability: Techniques and Design Methods for Constructing Long-lived, Heterogeneous Systems That Evolve over Time and Space While Remaining Interoperable
EmNets will often be embedded in long-lived physical structures (homes, office buildings, hospitals, wells, aqueducts, airplanes, roads, and so on) and thus must be long-lived themselves in order to be effective. To be long-lived, EmNets must be able to evolve, as it is very likely that the functionality required of them will change in some way, perhaps to something for which they were not originally designed. Further, heterogeneous EmNet components will have to interoperate with each other, as well as with various external devices to which they will connect. Achieving such interoperability over the lifetime of the EmNet and over the changing space in which the EmNet will be operating is an open research challenge. As discussed throughout the report, existing techniques and strategies for interoperability are not yet up to the many challenges posed by EmNets.
EmNets will typically operate in an unattended mode, wherein many actions must be taken without human intervention. Aspects of the environment may change, and elements may be moving into and out of the system in unanticipated ways without user assistance. Moreover, while day-to-day operations will need to occur autonomously, the system itself
may also have to evolve without human direction. Thus, both the normal operation as well as the system evolution of the EmNet need to be self-configuring. In addition, the operational details of EmNets are often hidden from casual users, and thus the evolution of the system needs to occur as transparently as possible so as not to be obtrusive.
The field of EmNets is developing rapidly but in an uncoordinated fashion. Because they were so badly needed, a number of EmNets have already been designed, built, and deployed, and many of them have come to us from fields other than computer science, such as aeronautics and systems engineering. If EmNets are not to risk becoming obsolete before they are deployed, system evolution and integration standards cannot really start from scratch but must allow the integration and evolution of existing legacy systems.
Accordingly, a research program is needed that will actively challenge EmNet research projects by requiring the integration of unanticipated elements into the research. These unanticipated elements might take the form of new devices, either tethered or mobile, or even legacy systems that could be of use to the overall system. The real aim of this requirement is to ensure that the framework developed for the EmNet is flexible enough to deal with new elements and new requirements. Left to their own schedules, researchers will design for what they foresee the future to be; it is important that this research describe ways to deal with a future that cannot be foreseen.
Integration of Technical, Social, Ethical, and Public Policy Issues: Fundamental Research into the Nontechnical Issues of EmNets, Especially Those Having to Do with the Ethical and Public Policy Issues Surrounding Privacy, Security, Reliability, Usability, and Safety
EmNets are capable of collecting, processing, and aggregating huge amounts of data. With the advent of large numbers of EmNets, the technological stage is set for unprecedented levels of real-time human monitoring. The sensors are cheap and unobtrusive, the computing and communications costs are very low, and there will be organizations with the resources and the motivation to deploy these systems. Thus, EmNets present a difficult challenge in terms of passive information disclosure. In the case of the Internet, privacy issues arise because as users browse for particular kinds of information they are often asked to divulge explicitly other kinds of information, or their clickstreams through and among sites produce information that sites may be storing without the user’s informed consent. In the case of EmNets, inadvertent, even unintentional revelations are much more likely. The monitoring these systems do will be
almost completely undetectable. The temptation to use such systems for law enforcement, productivity monitoring, consumer profiling, or in the name of safeguarding children from harm will be enormous. At the same time, we have already seen effects of information moving quickly around the Internet (for example, false rumors have had dramatic effects on the stock markets (Walsh, 2000)). EmNets as they have been described here have the potential for even greater and more far-reaching effects.
With respect to security, history has shown that computer systems will be attacked. Data will be stolen or compromised, system functionality and/or availability will be impaired, and the attacks will be incessant. EmNets will be very much at risk for such attacks, since they are deployed specifically to collect important information about the real world and may be capable of acting on it. The security facilities of, say, the Internet, are obviously inadequate. EmNets require much better resistance to malicious intrusions and much better means for detecting and reporting such attempts. These issues are not merely technical, however, and will need to be addressed at a procedural and public policy level as well. The committee believes that purely technical approaches will be insufficient and that policy and technical aspects should be coordinated in order to address these problems. Privacy, security, and ethical considerations need to be considered and incorporated early, during the design and development phases of these systems. These are areas in which inter-and multidisciplinary research efforts could pay large dividends.
The committee believes that the ethical concerns related to security and privacy—which drive legal and policy activity—require a fundamental research agenda. Some of that research will relate to technical mechanisms that can help to ensure authenticated use and proper accountability while safeguarding privacy. But, perhaps more importantly, it may be necessary to develop a new calculus of privacy to be able to evaluate how interactions between new elements will impinge on security and privacy. Users will need ways of comprehending how the aggregation of the information they are divulging to disparate sources can compromise their privacy (e.g., connecting automobile sensor logs to location sensing), and they will need to move beyond concerning themselves only with the security of a Web site’s credit card files.
While this report’s primary focus has been on a technological research agenda, the committee strongly recommends also examining the policy and social implications of EmNets and other kinds of information systems. How can the development of policy and technical mechanisms be coordinated to encourage realizing potential benefits from EmNets without paying avoidable societal costs? Research that relates technical, social, and policy issues is consistent with the Social, Economic, and Workforce (SEW) component of the federal Information Technology
Research and Development program. This recommendation echoes an earlier CSTB recommendation that networking research should have a component that looks at ethical, legal, and social implications, drawing inspiration from the ELSI component of the human genome initiative.8
Enabling Technologies: Ongoing Research into the Various Component and Enabling Technologies of EmNets
In Chapter 2 several fundamental enabling technologies for EmNets were discussed. As described there, research in these areas is still needed in order for the full potential of EmNets to be realized. Several specific issues are mentioned here, although it should be noted that each of these technologies could generate an entire research agenda on its own.
First, continuing research into building low-power processors is essential for ubiquitous, efficient EmNets. Exploring the conflict between power efficiency and flexible functionality raises a number of interesting research questions, and determining the best way to approach this problem is an open question. Continuing research is also needed into wireless communications and network architectures for short-range, low-power systems. Open questions remain about where to place communications in relation to computation and where storage should take place, as well as what appropriate media access control (MAC) or MAC-level protocols should be. Alternative power sources are needed that will satisfy the form factor, communications, and computational requirements of EmNets and their individual components. The use of techniques such as ultrawideband (UWB) communications for EmNet applications should also be explored.9
EmNets will require changes in software functionality and development as well. Upgradability, high availability, and the ability to work with new hardware are just a few of the issues that will need to be taken into consideration when developing software for EmNets. Morever, new and better tools for software development will be needed to effectively and efficiently build software for these systems. Geolocation will also need to be further explored. Determining whether assisted GPS is an optimal location technology for EmNets is an open research question. At the same time, alternative techniques such as acoustic signaling should be explored. Finally, further work in MEMS sensors is needed to develop
sensors that can be realized on the same chips as the electronics needed for control and communication.
STRUCTURING THE RESEARCH ENTERPRISE FOR EMNETS
Ensuring that the right kinds of research are conducted to advance the state of the art in EmNets will require changes in the way the nation’s research enterprise is organized. Academia and industry will both have important roles to play. Effective collaboration will be needed not only among industry, universities, and government, but also between IT researchers and researchers in other areas that will make use of EmNets (e.g., the health sciences, manufacturing, and defense). Explicit efforts will need to be made to put mechanisms in place for ensuring such collaboration.10 While past attempts to achieve similar goals met with mixed results, the pressing needs of EmNets demand redoubled efforts, drawing upon the lessons of history.
Research directions, such as those described in the preceding section, are important to articulate, but it is also how that research is conducted that will determine whether the necessary advances are made. In the case of EmNets, researchers will have to gain experience in building and deploying systems. Many of the properties that will need to be studied will emerge only when elements are deployed and begin to be combined and coordinated in ways not foreseen by their designers.
Research funding agencies must be ready to promote a long-term, comprehensive vision and ensure that the appropriate communication occurs between the members of all relevant communities. Building sharing inter- and multidisciplinary communities is essential in a critical research area like EmNets. Once established, these communities fuel research in both universities and industry and further development in industry. Experimental research (not necessarily separate from fundamental research) is key to advancing the EmNet agenda.11 This means building new systems, deploying them, evaluating them, and then redesigning or retuning the elements as well as the system as a whole. This is an iterative process, and many systems and elements will be thrown away along each cycle as new and better ideas and artifacts are developed.
Stimulating Interdisciplinary Research
Mechanisms will be needed to promote interdisciplinary approaches to research on EmNets, which tie computer science to other sciences and other disciplines in general. (See Box 6.1 for a discussion of what may be required when there is an increased emphasis on interdisciplinary and system-level approaches in educational environments.) Domain expertise found in disciplines such as biology, geophysics, chemistry, and medicine will allow the application of EmNets in a variety of areas. These disciplines and others can provide models that couple the world of the networked computer and the physical world and can help in investigations of the wider implications of EmNet society. A wide variety of application domains can serve as testbeds for EmNet ideas and concepts as well as bring richly interdisciplinary teams of researchers and scientists together. However, it is not simply a matter of bringing EmNet expertise to solve problems in the various sciences.
Interdisciplinary benefits will also flow in the other direction. It is clear that if EmNets are going to interface to the physical world, the engineers and computer scientists who will be developing EmNets will need to connect with those who understand the physical phenomena and all their manifestations and variations. These will include bioengineers, environmental engineers, mechanical engineers, nanotechnologists, earth scientists, and chemical engineers. Concepts from control theory and signal processing will need to be in the repertoire of every researcher.
Nor does the challenge end here, for the interdisciplinary net will need to be cast wider still, to bring concepts and techniques from even more distant disciplines, such as systems engineering, biological sciences, economics, and even sociology and political science. Each has a long tradition of trying to understand the aggregate behavior of systems that self-organize or that show coordination without centralized control. EmNets will be systems that are not open to centralized control in the same way that traditional computers or networks of IT machines have been. They will have to be self-regulating, self-configuring, and self-monitoring and will have a much higher degree of autonomy than previous systems, necessitated by the sheer number of devices that will be interconnected in many applications. Moreover, devices will be fielded that, because they will be deeply embedded in the environment or in larger artifacts such as vehicles or buildings, will have much longer lifetimes and will be upgraded by the addition of new elements rather than simple replacement. It is likely that much can be gained from looking at other disciplines to see what kinds of self-organization and decentralized controls have worked in other fields and whether any of the knowledge is applicable to EmNets. Such investigations could add many new pieces to the toolbox of EmNet research and development.
Increased emphasis on interdisciplinary and system-level approaches is crucial to moving forward in EmNet research. These two approaches are also the ones that require the most attention in the nation’s educational system. Related to them are four areas that are largely absent from engineering curricula today:
Most computer science and electrical engineering departments today are highly compartmentalized. Students are specializing in their studies at an earlier age and often come to higher education along a predetermined path that permits no forays into other disciplines. This tendency to be narrowly focused is often too limiting. Courses that look at the trade-offs between all the levels in the design of a complete system are rare. Furthermore, few institutions are able to couple traditional education with exposure to system prototyping because the technology is constantly evolving and the faculty have limited experience. System prototyping is an area ripe for collaboration with industry.
Interdisciplinary Educational Approaches
Interdisciplinary education is too often interpreted as intersubdisciplinary, since it is usually more expedient to think in terms of a single academic department. Students rarely work with students from other departments. Some successful examples come from closely related subdisciplines in engineering departments, but much more needs to be done in preparing for a world of EmNets.
Student design teams need to become broader. For example, the design of a new patient-monitoring and information system should involve students not only from medicine but also from public policy, law, and business, along with the computer science students who will actually write the code. The code they write—its organization as well as its function—may be deeply affected by their collaboration with students from these other disciplines. Electrical engineers developing new environmental sensor technologies, for example, would be well served by working not only with chemists but also with computer scientists, biologists, and other life scientists. This interaction will undoubtedly uncover new uses for the technologies as well as different, possibly much more efficient and/or effective approaches to solving the original problem.
Unfortunately, today’s highly specific courses must be taught by faculty from a single department and do not expose students to the rich fabric that interconnects all university disciplines. Graduate education does not correct this deficiency. In fact, it exacerbates the problem by demanding a deeper dive into one subdiscipline. Generalists are generally discouraged in most graduate programs. The
emphasis is on depth in a narrowly defined area. Few students are lucky enough to be involved with truly interdisciplinary research projects.
The challenges that lie ahead involve devising models for cross-department faculty collaboration, which is hampered today by antiquated models of teaching. Interdisciplinary teaching is rare, because academic institutions have yet to figure out a way to do accounting except at a departmental level. Finally, industry has a role to play in creating the kinds of educational programs needed for EmNets. By the very nature of the academic establishment, most faculty stop being practitioners for a large part of their careers. This is even more so in engineering than in other fields such as law or medicine. Involving leading industry practitioners in EmNet education is critically important to producing graduate students who think along multiple dimensions and view systems in the large, as integrated wholes rather than individually optimized elements.
The fact that components rather than systems are taught is an often-heard self-criticism of engineering faculty. But one person’s system is another’s component. So what is really meant by this? The fundamental difference is one of approach to a problem. Should the emphasis be on abstraction or analysis? Should reuse of modules be encouraged or everything be constructed from scratch? Are system integration issues of interoperability and testing given first-class status or are they afterthoughts?
The nation’s current educational system is ill equipped to teach design methodologies. Many perceive the topic as not difficult enough. Furthermore, it is a topic with which faculty have little or no direct experience. Yet, it is clearly a topic that will need much attention as we start to design EmNets, for they present a new framework distinct from that of more traditional systems. Without appropriate methodologies, formalizations, and abstractions it will not be possible to meet the challenge of graduating students at all levels who can function well in this new space. Most engineering disciplines could use courses in aspects of system design from evolution to manufacturing to safety. The focus today is too much on cost or size or power. Rarely are these issues considered in combination, and they are only a few of the many dimensions EmNet designers will need to face.
Current teaching methods are based on understanding components, or “design in the small.” There is a bias toward teaching students how to design from scratch rather than to reuse what is available. Many faculty members find it difficult to understand how students can complete a degree without knowing how to do every component on their own. However, this style of thinking has led to an overemphasis on design in the small and a lack of exposure to design for reuse and the reuse of designs.
Instead, students should be encouraged to learn not only how to comprehend and build mental models of how others’ components work but also how to design so that others can share their design artifacts. Currently, abstractions permit this
at lower levels (for example, logic gates and protocol stacks), but higher levels need to be used (for example, self-updating code and composable network services) if systems of the scale and complexity of EmNets are to be built. Fostering the development of formal models that support higher levels of abstraction and provide students with a curriculum that lets them build on others’ work while also providing building blocks for those coming after them is key to this endeavor.
Finally, one of the most important educational experiences is to work through the process of bringing together a system of many components. This step is crucial to understanding the value of design methodologies and abstractions. System design without the experience of integration is similar to writing code that is never debugged. The art of stepwise integration and debugging needs to be imparted to students as early as possible in their curriculum, and they should be repeatedly exposed to these issues throughout their education.
It is important to understand that the term “integration” is meant in the broadest possible sense. That is, it comprehends not only integration of the components but also the deployment (or integration) of the system into its intended operating environment. Any system will alter that environment and thus affect the assumptions that underlie its own design and development. The closure of that feedback loop is a fundamental lesson in the process of design that few students gain from today’s engineering education.
Because of their scope, EmNets offer a new opportunity for cooperation between academia and industry, both in the traditional channels of the computing industry and academic computer science departments and in new channels of interaction between a wider set of academic departments and computing and noncomputing industries, such as medical equipment manufacturers, environmental monitoring consultants, and resource management industries. The committee recognizes that fostering successful interdisciplinary and interinstitutional research is not easy. Encouraging such interdisciplinary and nontraditional collaborations will require the creation of new research venues and new incentives for industrial and academic partnerships. Educational institutions will need to be encouraged to create new centers for research that cross traditional departmental boundaries and ensure that research opportunities within these centers are funded and rewarded. Funding agencies will need to think “outside the box” about the kinds of collaborations they accept and
promote. New industrial partners will need to be approached, educated, and enlisted in the construction of new systems that solve problems not currently thought of as part of networks of computers.
WHAT CAN GOVERNMENT DO? RECOMMENDATIONS TO FEDERAL AGENCIES
The federal government has long been a strong supporter of broad-ranging research in information technology. While there have been numerous notable successes—indeed, whole industries have grown out of this funding12—fundamental research in information technology is far from complete. This is clearly seen in the context of EmNets. For the most part, EmNets are currently deployed in application-specific, highly engineered contexts. It is essential to develop mechanisms, algorithms, and models that are broadly applicable and reusable to gain experience and confidence with various approaches over time. Similarly, a base of trained technical personnel is needed who understand how to design, develop, and implement these systems. While it is powerful and compelling to demonstrate the concepts and see the potential in various prototypes, such demonstrations alone will not develop the discipline and the techniques to fulfill the vision outlined in this report. Long-lived research programs are essential so that the deeper, harder issues can be addressed and a set of well-understood, characterizable primitives developed for use across many application instances—this is where university research becomes crucial for complementing the more directed and sometimes narrower scope and shorter-term focus of industry.
Federal funding for research guides the focus of the university research community and influences not only what is accomplished there but also what is accomplished in industry. Such funding can cause industry to take a broader perspective and produce more flexible technology for users in the federal government and elsewhere than it would if left strictly to market forces. Collaboration is necessary between industry and academia as the science of EmNets is developed. Today, many university projects are too close to product development, with the lure of start-ups having done much to push things in this direction. Models for joint investigation, fostered by appropriately targeted federal funding, should be renewed if the research community and society are to reap the benefits
of a full collaboration. To that end, the committee next describes several ways in which the Defense Advanced Research Projects Agency (DARPA), the National Institute of Standards and Technology (NIST), and the National Science Foundation (NSF) could facilitate research in these areas. It also makes several recommendations to various federal agencies regarding effective sponsorship and support of EmNet-related research.
Recommendations to the Defense Advanced Research Projects Agency
DARPA has already invested in EmNet-related technologies, but it has only scratched the surface of what will be necessary to advance this critical technology. Both its Information Technology Office (ITO) and its Microelectronics Technology Office (MTO) have developed programs that relate to EmNets. It is now time to build on the past successes and present efforts13 and to broaden and deepen the work in this area. A multifaceted program or set of programs is needed that will pursue the core computer science and information technology issues that have been raised throughout this report. As described previously, narrowly focused solutions and small-scale programs are a good and even essential start, but they are not up to the gigantic task of developing reusable, generalizable, characterizable, and robust techniques for designing, implementing, deploying, and operating large-scale, robust EmNets. It is time to build on these endeavors and turn to systems work that will require extensive breadth and depth in order to be successful.
Publicly funded research is needed to drive innovation that is of sufficient scope—that is, that covers predictability, adaptability, survivability, system monitoring, and so on—and addresses externalities such as interoperability, safety, and upgradability. The development of robust EmNet technology will require the research community to rethink the fundamentals of information technology and the design of computer and communications systems. First and foremost it calls for a systems approach in which design, programming, and control focus on systems composed of massive numbers of networked components and not on optimization of individual or small numbers of elements. A single, isolated,
short-lived research program will not suffice to address the scope and depth of the problems that must be addressed to realize scalable, robust, and usable EmNets. DARPA should aggressively pursue multiple programs that build upon and interact with one another and with some of the seed programs that have already begun to explore related areas. These seed programs—SensIT is one—have made important initial contributions. It is in part their successful initial forays that now allow the committee to articulate a full-fledged research agenda. However, as mentioned before, they were not of the scale, duration, or scope needed to address DARPA’s critical medium- and long-term needs for robust, scalable EmNet systems technologies, and DARPA should now encourage the development of multiple programs that build upon and interact with one another. To truly harness the power of EmNet systems, DARPA should manage these programs in a way that fosters their interaction and creates and builds on conceptual overlaps. The committee emphasizes the need for intellectual collaboration and communication as opposed to requiring prototypes or deliverables from each project for use by one or more of the other projects. There is much to be gained by understanding and exploiting the conceptual commonalities across networked embedded control systems, ad hoc sensor networks, low power design, and smart fabric. And there is much to be lost if such collaborations fail to materialize.
Making progress in an area as large as and, in many ways, as radical as EmNets requires sustained support for research along with a careful rethinking of how best to organize, communicate, and develop the work over the long term. EmNets present an opportunity to continue progress in critical areas of information technology research as well as to discover and advance new capabilities. A long-term research agenda that begins to address these challenges in parallel, while promoting cross-collaboration and interdisciplinary, interprogram work where appropriate, will have tremendous impact. It should have sufficient longevity to explore multiple approaches without insisting on preaward or preresearch agreement on the general architecture and infrastructure. To this end, two recommendations are given below, along with a (by no means comprehensive or canonical) list of possible DARPA programs in this area.
Recommendation 1. The Information Technology Office of the Defense Advanced Research Projects Agency should revise both the substance and process of its EmNet-related programs to better address the research needs identified in this report.
DARPA’s Information Technology Office (ITO) took the lead in early research on sensor networks. However, there are several ways ITO’s programs could more fully address the research needs explicated in this
report. Field demonstrations are clearly critical to DARPA, and such demonstrations should continue. However, the committee suggests that early in a technology’s development, research dollars are better spent on exploration of the design space and experimental exploration than on field demonstrations of particular point solutions. Such demonstrations can crowd out more systematic investigations and higher-risk investigations and tend to place too much emphasis on early system integration and convergence to single approaches. Carefully crafted experimental work, on the other hand, can promote real system development and use in a context that provides invaluable feedback to researchers and developers. While it is important for universities to build prototypes, it is crucial to remember that these prototypes are built not for future product development, as are those built by industry, but to understand better the problems of the application. That deeper and more focused understanding is what brings about innovative solutions to problems by deepening scientific understanding (determining, for example, formal models and appropriate abstraction layers). Experimental projects might even involve the definition of interfaces and integration over time without, however, being limited by the constraints of time-sensitive demonstrations. After some period of time, contractors (i.e., industry) should be involved in developing demonstration prototypes and should share their experiences with researchers.
The committee recommends that DARPA focus its efforts on four technical areas in order to realize EmNet technology that is robust, scalable, and widely applicable across Department of Defense needs, both on the battlefield and off (e.g., logistics). These areas are described in Box 6.2. Some of these topics are being addressed by individual principal investigators who are or have been funded under one of DARPA’s existing EmNet-related programs, such as Ubiquitous Computing14 (part of this program focuses on the notion that users do not interact with the computing devices themselves but with the services they provide) and SensIT15 (the emphasis in this program is creating connections between the physical world and computers by developing the software for networked sensors). Box 6.3 describes more of ITO’s current and recent programs in this area. However, the topics addressed by each of these programs deserve and require more exploratory, broader-based investigation. The programs suggested in Box 6.2 are far from exhaustive, but they could serve as the beginning pieces of a much larger systematic effort to address the issues raised in the box.
For more information, see <http://www.darpa.mil/ito/research/uc/index.html>.
For more information, see <http://www.darpa.mil/ito/research/sensit/index.html>.
Designing for Predictability, Reliability, and Safety
As more and more technology is employed in support of mission-critical operations, the inadequacy of system predictability and diagnosability is posing tremendous risks. EmNets intensify these inadequacies, because users will typically interface with the object in which the EmNets is embedded rather than with the system itself. A program is needed to develop abstractions and models that allow users to understand and reason about variable system conditions and failures. Rather than developing models for safety, reliability, and predictability separately, it is critically important to develop models that encompass all three and that address the trade-offs that will be necessary among them. Further, it is increasingly important to build systems with quantifiable (in some cases, provable) properties such as scoping or isolation of system behaviors.
Collaborative Signal Processing
While DARPA has initiated some programs in the area of EmNets that apply to sensor networks, there is a particular need to engage the signal processing community in the development of distributed collaborative signal processing across multiple sensory modalities. Existing programs in these areas require renewed emphasis and support.
Multi-scale Location-aware Systems
Technology has been and is being developed to support particular geolocation techniques. However, many forms of geolocation that are related to proximity and logical location must be integrated into EmNets. There should be a program promoting system technology that exploits multiscale location and involves approaches that will work through a variety of media, including RF, acoustics, and imaging. The program should also explore the difference between infrastructural and noninfrastructural (more ad hoc) approaches.
Interoperability over Time and Space
EmNets will be embedded in our infrastructure and therefore will have lifetimes as long as that of the infrastructure. At the same time, new devices will continually be introduced into the overall system. A program that addresses the challenges of integration and interoperability with new devices over long system lifetimes and changing expectations is needed. It should emphasize research in how to handle legacy devices (for example, how to decommission them while they are deeply embedded). Further, such a program should incorporate the notion that units of interoperability vary: A single device may need to interoperate with other devices, or a cluster of devices may need to interoperate as a unit with other clusters of devices.
Networked Embedded Software Technology (NEST)
In this project, DARPA is seeking novel approaches to the design and implementation of software for networked embedded systems. The coordinated operation of distributed embedded systems makes embedding, distribution, and coordination the fundamental technical challenge for embedded software. The goal of the NEST program is to enable fine-grained fusion of physical and information processes.
Sensor Information Technology (SensIT)
The goal of the SensIT program is to create the binding between the physical world and cyberspace. SensIT is founded on the concept of a networked system of cheap, pervasive devices that combine multiple sensor types, reprogrammable processors, and wireless communication.
The goal of the Ubiquitous Computing program is to create a post-PC era of computing in which a scarce resource—human attention—is conserved in an environment where computing functionality is embedded in physical devices that are widely distributed. In this environment, users do not interact with any particular computing device but rather with the functionality and services offered by the set of devices at hand.
Recommendation 2. The Defense Advanced Research Projects Agency should encourage greater collaboration between its Information Technology Office (ITO) and its Microelectronics Technology Office (MTO) to enable greater experimentation.
There is an opportunity to take advantage of collaborations between ITO and MTO by enabling experimental EmNet projects with real state-of-the-art sensors and even actuators. MTO-funded research has brought significant advances in MEMS technology, but that research has not yet emphasized the system-level aspects of MEMS. (See Box 6.4 for recent work in EmNet-related areas in DARPA’s MTO and its Advanced Tech-
Model-based Integration of Embedded Software
The goal of this project is to create a new generation of system software that is highly customizable and responsive to the needs of various application domains and to the constraints of embedded systems.
The goal of the Power-aware Computing/Communication project is to enable the intelligent management of energy and energy distribution, providing the minimum power necessary to complete a given task.
Adaptive Computing Systems
The Adaptive Computing Systems program was designed to create unprecedented capabilities for the dynamic adaptation of information systems to a changing environment. It explores redefining the traditional hardware/software boundary to enable the rapid realization of algorithm-specific hardware architectures on a low-cost COTS technology base.
The Embeddable Systems program focuses on leveraging and extending the commercial scalable computing technology base to support defense embedded-computing applications.
Software for Distributed Robotics
The goal of this project is to develop software for the employment and control of large numbers of small, distributed, mobile robots in order to achieve large-scale results from many small-scale robots.
nology Office (ATO).) The idea is to apply well-understood MEMS techniques to produce several types of sensor/actuators that can be integrated into EmNet prototypes by the research community and allow for more realistic experimentation with a range of physically coupled systems. These might take several forms. Examples include a chemical sensor that could be used in experimental monitoring systems, a computational fabric that has a mixture of pressure and temperature sensors, and tension-varying actuators that would enable experimenting with how to control EmNets of this type.
The research community could define standard interfaces to these
The DARPA Distributed Robotics program seeks to develop revolutionary approaches to extremely small robots, reconfigurable robots, systems of robots, biologically inspired designs, and innovative methods of robot control. The program focuses on individual robots that are less than 5 cm in any dimension.
Microelectromechanical Systems (MEMS)
The primary goal of the DARPA MEMS program is to develop the technology to merge sensing, actuating, and computing in order to realize new systems that bring enhanced levels of perception, control, and performance to weapons systems and battlefield environments.
Microoptoelectromechanical Systems (MOEMS)
The primary goal of the MOEMS program is to develop the technology to merge sensing, actuation, and computing in order to realize new systems that bring enhanced levels of perception, control, and performance to military and commercial systems.
The Smart Modules program is developing and demonstrating novel ways of combining sensors, microprocessors, and communications in lightweight, low-power, modular packages that offer warfighters and small fighting units new methods to enhance their situational awareness and effectively control their resources on the battlefield.
Future Combat Systems Communications
The goal of this program is to produce communications technology for ad hoc networks that can operate under severe operational constraints, such as a hostile electromagnetic environment. These mobile networks will have both airborne and terrestrial platforms deployed in an autonomous fashion to provide needed coverage on an ad hoc basis.
Global Mobile Information Systems (GloMo)
The goal of the GloMo project was to make the environment a high priority in the defense information infrastructure, providing user-friendly connectivity and access to services for wireless mobile users.
devices and enable relatively inexpensive prototyping in a widespread manner. Such technologies would provide the academic research community, in particular, with the kinds of artifacts it will need to better explore applications of MEMS technology to EmNets and the system-level issues that result.
Recommendations to the National Institute of Standards and Technology
NIST, and in particular its Information Technology Lab, has worked in a variety of areas to help make information technology more secure, more reliable, more usable, and more interoperable. All of these characteristics are, as has been described, crucial to current and future EmNet-related technologies. NIST has played a valuable role in promoting standardization and acting as a verification agent (see Box 6.5 for information on EmNet-related NIST programs). In this role, NIST establishes trust in techniques and mechanisms by establishing testing and evaluation standards. Many applications and components of EmNets will require verification, and NIST is in an excellent position to act as arbiter between developer and user.
NIST has already begun to play a role in wireless interference and associated power and frequency standardization. This effort will become even more critical as more wireless devices are deployed at greater densities.16 New applications of EmNets will call for entirely new metrics for evaluation (such as system lifetime and system manageability or instrumentation). A wide range of standardization efforts will be launched as an offshoot of EmNet activities, including sensor, actuator, wireless, and cross-system interactions.
NIST is in an excellent position to foster interaction by devising the appropriate metrics for measuring the effectiveness of EmNet elements as well as the requirements for performance and quality of service for the more abstract services that will be built upon those elements. In addition to metrics, NIST can also act as a collector of and repository for experimental data. There is a growing gap in access to critical evaluation data. This is already evidenced in the case of the Internet. Unlike in the early days of computing, when most researchers could manage to measure the performance of their own computing equipment, today a national- or even a global-scale infrastructure is required for collecting data-traffic information. Such an infrastructure is accessible to only a very few large
The NIST Smart Space Laboratory
Smart spaces are work or home environments containing embedded computers, information appliances, and multimodal sensors. NIST’s goal is to address the measurement, standards, and interoperability challenges that must be met as tools for these environments evolve in industrial R&D laboratories worldwide. NIST is also working to develop industrial partnerships and is sponsoring workshops with DARPA and NSF in this area.
Networking for Smart Spaces
This project explores the use of Java, Jini, and multicast technology in conjunction with wireless systems such as Bluetooth and HomeRF as a networking foundation for pervasive computing or smart spaces.
The Aroma Project
The goals of the Aroma project are to help research, test, measure, and standardize pervasive computing technology by, among other things, measuring the resource requirements and performance of emerging pervasive computing software and networking technologies; developing software tools for testing, measuring, and diagnosing pervasive software and networks; and creating standard abstractions and models for developers.
companies. Expanding access to this data by more researchers is an important role for a government agency.
The committee believes that NIST also has a particularly critical role to play in this realm as the agency that establishes confidence in information systems. NIST is seen as an outside observer that can provide objective services and analysis. It has an important role in the standards-development process, allowing the work done in industry to be illuminated in a fair and open fashion. As this report has emphasized, interoperability for EmNets will be very important, and standards will be needed for such interoperability. Given that many of the standards in this arena are likely to arrive as de facto rather than de jure standards, NIST can provide an objective analysis of them and reduce barriers to entry with reference implementations of the technology itself and/or reference implementations of conformance testing tools. More specifically, NIST, through activities such as its Aroma Project,17 which focuses on testing,
For more information, see <http://www.nist.gov/aroma/>.
measuring, and standardizing pervasive computing technology, should play a significant role in the two areas as EmNets become ever more widespread.
Recommendation 3. The National Institute of Standards and Technology should develop and provide reference implementations in order to promote open standards for interconnectivity architectures. It will be important to promote open standards in the area and promote system development using commercial components by making public domain device drivers available.
Recommendation 4. The National Institute of Standards and Technology should develop methodologies for testing and simulating EmNets in light of the diverse and dynamic conditions of deployment. Comprehensive simulation models and testing methodologies for EmNets will be necessary to ensure interoperable, reliable, and predictable systems. In particular, the development of methodologies for testing specification and interoperability conformance will be useful.
In the process of these endeavors, NIST can play a key role in data collection and dissemination of EmNet-related information for use by the larger research and development community.
Recommendations to the National Science Foundation
The National Science Foundation (NSF) has a strong track record in promoting multidisciplinary research and integrated research and education programs. More recently, it has been increasing its support for integrated systems projects—for example, the Information Technology Research (ITR) program. All three areas—multidisciplinary research, integration of research and education, and integrated systems approaches—will be of great importance in the support of EmNet-related research projects, and all of them—in particular, systems-oriented work—should be aggressively pursued and include cross-divisional efforts where necessary. Specific recommendations for NSF are below.
Recommendation 5. The National Science Foundation should continue to expand mechanisms for encouraging systems-oriented, multi-investigator, collaborative, multidisciplinary research on EmNets.
NSF is funding work in several areas related to EmNets (see Box 6.6). Much of this work continues to be done by a single principal investigator (and graduate students) operating on a small budget. As noted in this
Scalable Information Infrastructure and Pervasive Computing
NSF is supporting work in scalability, security, privacy, sensors and sensor networks, and tetherfree networking and communications in this program. Its goal is to advance the technical infrastructure to support human-to-human, human-to-computer, and computer-to-computer remote communication.
Wireless Information Technology and Networks
This program funds research to provide a foundation for designing high-information-capacity wireless communication systems for full mobility. Such design will require synergistic, multidisciplinary research efforts encompassing a breadth of communications functions from the physical through application layers.
Electronics, Photonics, and Device Technologies
This program funds research in the areas of micro- and nanoscale devices, components, and materials, advanced methods of design, modeling, and simulation of such devices and components, and improved techniques for processing, fabrication, and manufacturing.
report, research on EmNets will require that such single investigator research be complemented by collaborative experimental research that brings together researchers from different disciplines to focus on a common problem. Had this report been written several years ago, it would have recommended that NSF move toward larger-scale, experimentally driven, risk-taking research. NSF’s ITR program appears to be doing just that. ITR also reinforces attention to the social and economic dimensions of information systems. This program, or others like it, could serve as a useful vehicle for pursuing some of the topics pinpointed in this report. The key to achieving successful multidisciplinary research is not just a matter of funding levels. A flexible process is required that can incorporate perspectives from a broad range of relevant disciplines.
Recommendation 6. The National Science Foundation should develop programs that support graduate and undergraduate multidisciplinary educational programs.
With respect to education (see Box 6.1), NSF could take the lead in tackling institutional barriers to interdisciplinary and broad systems-based work. NSF has a history of encouraging interdisciplinary programs and could provide venues for such work to be explored (as is being done in the ITR programs) as well as foster and fund joint graduate programs or joint curriculum endeavors. One way to do this would be to provide incentives to programs that successfully cross disciplinary boundaries. For example, faculty working on interdisciplinary research often have difficulty securing institutional support for work deemed outside the scope of their home department. A program that removed this drawback by providing funding for such work could stimulate interdisciplinary research and course material in colleges and universities. Another way would be to expand the Graduate Fellowship Program to support more interdisciplinary proposals. Suitable evaluations of proposals would be needed to implement this recommendation.
Recommendations to Other Federal Agencies
The National Aeronautics and Space Administration (NASA) and the Department of Energy (DOE) were two of the earliest innovators and adopters of EmNets. While NASA and DOE application domains can be quite specialized, two things are clear: The computer science community would benefit from hearing of and seeing this earlier (and contemporary) work, and NASA and DOE themselves would benefit from the more general pursuit of this technology by the broader computer science community. Both agencies have long histories in systems engineering as well as in computer science and so could serve as a useful bridge between various communities, especially regarding the development of EmNets. NASA, for example, has a strong interest in safety and reliability, and DOE has long been involved in reliability issues. Their expertise, when applicable, could be shared with others in related research areas; in addition, the two agencies would benefit from the generalizations that the broader research community could provide. More explicit cooperation and communication would be beneficial to everyone and would greatly advance the field.
The agencies with needs for EmNets should together promote expanded experimental research with a shared, experimental systems infrastructure. The committee expects that coordination needs could be supported by the various organizations and groups associated with federal information technology research and development.18 Open-platform sys-
tems of various scales, low-power components and the software drivers for these components, debugging techniques and software, traffic generators—all can be shared across research programs when applicable, avoiding inefficient redundancy in those parts of the system where there is more certainty. The research communities should combine their efforts in creating enabling components, such as a range of MEMS-based sensors and actuators that are packaged in such a way as to be easily integrated into experimental EmNet systems. This would enable experimentation with EmNets in environmental and biological monitoring applications, for example, that are relevant to a variety of agencies, such as the Environmental Protection Agency, the Federal Aviation Administration, the National Institutes of Health, the National Oceanic and Atmospheric Administration, DOE, and NASA, as well as research groups working in these areas. Cross-collaboration and communication and the development of general enabling components will be essential for broad-ranging experimental work with EmNet systems.
EmNets present exciting new challenges in information technology, posing fundamental research questions while being applicable to a broad range of problem domains and research disciplines. Unfortunately, progress in this area will probably be confined to domain- and application-specific systems unless a concerted, comprehensive effort is made to broaden and deepen the research endeavor. It is unlikely that such a broad-based, widely applicable research agenda will be undertaken by industry alone. While systems can be built individually, the accumulated understanding will be insufficient without fundamental work promoted and supported by federal funding agencies. The technology would also be much more expensive, only narrowly applicable, and far less extensible and robust. Long-term, forward-thinking, and broad-ranging research programs are crucial to achieve a deep understanding of EmNet impacts on society and of how to design and develop these systems.
Computer Science and Telecommunications Board (CSTB), National Research Council. 1994a. Academic Careers for Experimental Computer Scientists and Engineers. Washington, D.C.: National Academy Press.
CSTB, National Research Council. 1994b. Realizing the Information Future; The Internet and Beyond. Washington, D.C.: National Academy Press.
CSTB, National Research Council. 1995. Evolving the High Performance Computing and Communications Initiative to Support the Nation’s Information Infrastructure. Washington, D.C.: National Academy Press
CSTB, National Research Council. 1999. Funding a Revolution: Government Support for Computing Research. Washington, D.C.: National Academy Press.
CSTB, National Research Council. 2000. Making IT Better: Expanding Information Technology Research to Meet Society’s Needs. Washington, D.C.: National Academy Press.
CSTB, National Research Council. 2001. The Internet’s Coming of Age. Washington, D.C.: National Academy Press.
Walsh, Sharon. 2000. “Feds make arrest in Internet hoax case.” The Standard, August 31. Available online at <http://www.thestandard.com/article/display/0,1151,18153,00.html>.