Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 11
2007-2008 Assessment of the Army Research Laboratory 2 Computational and Information Sciences Directorate INTRODUCTION The Computational and Information Sciences Directorate (CISD) was reviewed as a whole by the Panel on Digitization and Communications Science of the Army Research Laboratory Technical Assessment Board (ARLTAB) during August 21-23, 2007, and July 29-31, 2008. In addition, subgroups of the panel reviewed meteorology-related work on November 5, 2007, and work on high-performance computing (HPC) on May 30, 2008. The reviews consisted of overviews by directorate and division management, presentations on a subset of current projects, poster sessions during which project leads were available, and laboratory tours. As of July 2008, CISD has grown to four research divisions: Advanced Computing and Computational Sciences Division (AC&CSD), Battlefield Environment Division (BED), Information Sciences Division (ISD), and Network Science Division (NSD). It also includes one infrastructure division, Information Technology, which serves all of ARL through its computing hardware, software, and staff. CISD is responsible for a continuing Collaborative Technology Alliance (CTA) on Communications and Networks and a continuing International Technology Alliance (ITA) on Network and Information Sciences. A new Advanced Decision Architectures CTA is co-managed by CISD with funding from the Human Research and Engineering Directorate (HRED). CISD’s expressed mission is to create, exploit, and harvest innovative technologies to enable knowledge superiority for the warfighter through advanced computing, network and communications sciences, information assurance techniques, and battlespace environment sensing and modeling. To carry out this mission, CISD performs research for the following purposes: To advance computational sciences and HPC technologies in support of Army systems; To perform atmospheric dynamics sensing and modeling for use in battlefield applications;
OCR for page 12
2007-2008 Assessment of the Army Research Laboratory To develop techniques for battlefield information fusion and processing, language translation, and autonomous agent control; and To develop self-configuring wireless network technologies that enable secure, scalable, energy-efficient, and survivable tactical networks. Tables A.1 and A.2 in Appendix A respectively characterize the funding profile and the staffing profile for CISD. CHANGES SINCE THE PREVIOUS REVIEW Since the last documented review (for the 2005-2006 period),1 several changes have affected CISD’s research activities. The first of these was the major reorganization in 2008 that increased the number of divisions from three to four. Only BED remained unchanged. The new addition was the Network Science Division, formed from assets in the prior Computer and Communication Sciences Division and the High Performance Computing Division. This new division, NSD, grew out of the recognition that networking sciences at several levels had become key to many Army future needs, especially the mobile ad hoc networks (MANETs) expected to be found in profusion on future battlefields. NSD’s charter emphasizes technologies that enhance tactical communications and networking capabilities both with warfighters and with sensor networks; methodologies to analyze, model, design, predict, and control the performance of such networks; and system architectures and algorithms to recognize and react to intrusion-detection events in such networks. The former Computer and Communication Sciences Division was reformulated into the Information Sciences Division, with a charter primarily focused on fusing timely information from all relevant sources for the warfighter in real time. In addition, the remaining assets of the former High Performance Computing Division were reformulated into the Advanced Computing and Computational Sciences Division, with a charter focused on using advanced computational sciences and high-performance computational resources. This division still manages or oversees two supercomputer facilities: the DoD Supercomputing Resource Center in Aberdeen, Maryland, and the Army High Performance Computing Research Center (AHPCRC), which has moved from Minnesota to California and is now directed by a team at Stanford University. Along with this reorganization, the Communications and Networks CTA continues to advance survivable and secure information communication and processing over wireless mobile networks. Also, the new International Technology Alliance on Network and Information Sciences was formed, involving participation from institutions in both the United States and Great Britain. The focus of this ITA is on managing end-to-end information flows in support of coalition decision making. The three newly announced CTAs for fiscal year (FY) 2009 in Robotics, Cognition and Neuroergonomics, and Network Science will all directly relate to CISD activities. In line with Army Research Laboratory (ARL) initiatives, CISD has also established a new Mobile Network Modeling Institute and has continued investments through the DoD High Performance Computing Modernization program. Moreover, CISD has continued investments through the Small Business Innovation Research (SBIR) program to bring in new technologies from emerging high-technology companies. 1 National Research Council, 2005-2006 Assessment of the Army Research Laboratory, Washington, D.C.: The National Academies Press, 2007.
OCR for page 13
2007-2008 Assessment of the Army Research Laboratory ACCOMPLISHMENTS AND ADVANCEMENTS Since the 2005-2006 assessment, the Board has seen achievements in the Computational and Information Sciences Directorate in three areas: continuing advances that have been made in key projects begun in the past, new projects that have come to fruition in the past 2 years, and a combination of reorganizations and new initiatives that are focused on future needs. Each is discussed below. Continuing Advances The 2005-2006 assessment report documented several areas of research that have continued to demonstrate significant advances over the past 2 years: Machine translation of foreign languages; Atmospheric acoustics, radio-frequency (RF), and optical propagation in battlefield environments; and Modeling of surface-level weather, especially wind. The machine translation work continues to demonstrate leadership in ways that directly aid the warfighter on today’s battlefields and sets the stage to provide rapid help as needed when new fronts emerge. The advances reported in 2006 were centered on porting text and speech translation engines to laptops and personal digital assistants (PDAs) for field deployment, and a “best-of-breed” testing and evaluation procedure for conducting relevant bake-offs of emerging new Machine Translation (MT) programs from ARL and outside organizations (both industrial and academic). The work in machine translation of foreign languages since then remains an archetype for goal-directed research with high value and frequent spin-offs. The big change since 2006 has been an increased emphasis not just on testing and evaluation but on actual translation, and doing so in the context of the kinds of work flows being experienced in the field. Thus, recent work has focused on handling not only voice but text. The goal for the latter is to help automate the processing of vast stacks of newly uncovered random documents not only by translating but by annotating them in ways that allow rapid key word and key phase searches after the fact. New languages beyond Arabic are receiving significant focus, from Urdu (in which there are at least limited amounts of training material) to languages based on Swahili or Hausa (in which there are often few or no training resources). The technology being developed at ARL leverages directly the outstanding testing and evaluation procedures described in prior ARLTAB reviews. In particular, the use of hybrid machine translation schemes that apply multiple different MT programs to the same documents in both serial and parallel combinations seems to be yielding better results than are observed for any one approach in isolation. ARL’s ability to realistically compare and fairly rank different MT programs is key to constructing, evaluating, and deploying such combinations. The other laudable aspects of this machine translation work are the significant amount of collaboration involved and the strikingly effective ways in which this collaboration is used. Universities, government laboratories, and industry are all integral parts of the mix, ensuring that ARL understands the current state of the art and how to combine multiple techniques into real deliverables for emerging applications on the battlefield. Cooperative Research and Development Agreements (CRADAs) have been developed with multiple partners, and joint work has been performed with the National Institute of Standards and Technology.
OCR for page 14
2007-2008 Assessment of the Army Research Laboratory Another area of clear and continuing outstanding progress in CISD is in selected areas of atmospheric acoustics and RF propagation in battlefield environments. The Board’s 2005-2006 assessment report highlighted outstanding research in developing and evaluating acoustic propagation models that incorporate environmental effects, both natural and human-made, on acoustic signatures, and in developing remote sensing techniques for use in a range of environments, from open deserts, to rugged terrain, to urban environments, and in a variety of weather conditions. Urban environments have grown in importance to many current Army operations, and an understanding of how acoustics are propagated in such environments can provide significant tactical advantages. Since 2006, ARL has continued this work, with growing emphasis not only on new sensor platforms such as very lightweight unmanned aerial vehicles (UAVs) and tethered aerostat balloons, but also on atmospheric effects such as temperature inversions and low-level wind jets and shears. Some of this work, such as on aerostat systems, had clearly identified new sources of noise that must be corrected for and had developed new near-Earth models to help in the analysis. The results deserve wider dissemination within the DoD, especially as the range of sensor platforms and the complexity of the environments continue to increase. ARL facilities and test ranges used in such efforts, especially at White Sands, New Mexico, remain unique and continue to contribute significantly to the relevance of the work. A related area of continuing high-quality work is in atmospheric optical propagation as performed in the ARL Intelligent Optics Laboratory, now part of ISD. The previous report commented on both the quality of the laboratory facilities and the way in which such facilities were being used. This has continued over the past 2 years, with a focus on developing optical systems for high-energy laser directed-energy applications such as targeting, atmospheric imaging, and communications. The key feature of such work is in adapting, in real time, multiple beams focused on the same target over long distances in the atmosphere, in order to achieve effects that would be the equivalent of those from a single, coherent, higher-energy beam. Real-world demonstrations have been conducted both at relatively short range (a few kilometers on the ARL campus), and much longer range (between mountains in Hawaii). Finally, work at constructing microscale wind models for complex terrains, especially where turbulence may be present, continues to exhibit very high quality, with attention paid to both computational efficiency (the desire to apply many such models to the battlefield) and verification against measurements from the White Sands, New Mexico, range. New Advances In addition to the continuing activities discussed above, several additional research thrusts have borne fruit over the past 2 years. In terms of detection systems, continued research is leading to new mechanisms that can possibly identify the presence of particles, such as chemical and biological, in aerosols and the atmosphere. Both fluorescence spectra and optical scattering properties have been studied, with both experimental and modeling efforts. As with prior detector work that has led to actual deployments, the fluorescent spectral work is backed up by good experimental techniques and excellent test facilities, and the optical scattering modeling is leading to novel predictive techniques that have been well received in reputable scientific venues, both journals and conferences. Focus within the aerosol research program might beneficially be shifted from particle identification to volumetric presence, an area that seems to be of extreme interest in the case of biohazards. Applications resulting from such a switch might expand from the current identification to include concentration, harmfulness, and the development of removal scenarios. A second suggestion would be to see how these efforts dovetail into other ongoing efforts outside ARL, such as the Department of Defense’s National Signatures Program.
OCR for page 15
2007-2008 Assessment of the Army Research Laboratory Compact LIDAR systems (radar using light) have also been a focus of ARL activity that appears to be yielding potentially fieldable systems for providing tactical meteorological data in real time. The approach taken by the BED seems to be a good one. Partnering with industry through an SBIR grant has resulted in a 100 times reduction in weight of the LIDAR system—from 3,000 lb to 30 lb. Further work is focused on making the associated computations efficient enough for execution on embeddable computing systems. The reasons for focusing on this work are that embedded systems will soon have the needed computational power (through multicore), and ARL will be able to further reduce the size, weight, and power consumption of the LIDAR system. Another new sensing effort that builds on existing expertise in atmospherics, optics, sensors, and computing is the development and experimental analysis of algorithms that use polarized thermal imaging to enhance targeting and tracking, that counter deceptive techniques such as camouflage and decoys, and that suppress background clutter and highlight the location of a target within a thermal image. This was an area suggested in the 2005-2006 assessment report. A focus on how to use a new lighter-weight sensor obtained through an SBIR collaboration may prove useful in developing complete packages for important next-generation applications such as the detection of improvised explosive devices (IEDs). By providing a more careful analysis of alternative technologies, such longer-range efforts are particularly valuable in supplementing other government-sponsored efforts that rely more on quickly deployable systems. Preliminary data indicate some potential for novel capabilities, but a more rigorous measurement program, coupled with enhanced algorithm development as well as an understanding of emerging and alternate sensors, will be needed to see this effort through to completion and handoff for system deployment. An example of an effort in more basic science that seems poised to improve a range of Army systems focuses on increasing the understanding of atmospheric turbulence in wind, moisture, and temperature, especially near Earth’s surface. This is a very important problem, especially as more and more Army activities occur in urban settings with complex wind paths. CISD is tackling the problem with the right balance of theory, modeling, and experimentation. The results should be applicable in the future for a variety of uses with respect to flying small UAVs, to plume dispersion from rotorcraft, to predicting the movement of chemical-biological clouds, and even to wildfires in both urban and wilderness terrains. An example of an outstanding engineering effort is the development of the “Blue Radio,” a small, wireless network interface card that is designed from the bottom up by ARL as a demonstration platform for implementing sensor networks, particularly ones that will be placed randomly on the ground and thus must rely heavily on surface wave propagation rather than on free-space propagation. This appears to be an excellent platform on which to build an experimental program because of its success in being deployed, its sophistication for addressing Army-specific issues (relative to commercial candidates), and the availability of local radio expertise. It also appears that this effort may represent one of the very few if not the lone remaining site for expertise in such radios in the Army. Unfortunately, there seems to be neither a well-articulated path forward for exploiting the Blue Radio to the extent possible, nor a serious attempt to compare and contrast it with other sensor radio projects such as that developed at the University of California, Berkeley. A deliberate series of experiments based on this platform could be used to develop a methodology for validation and verification of the simulation, emulation, and theoretical efforts related to sensor radios. While this is only a specific example relevant to sensor networks, this methodology may very well form a strong base for analyzing other radio systems. An important advantage of using a captive radio system such as the Blue Radio would be the opportunity to develop and demonstrate radio networks that also could host advanced applications; such applications could demonstrate the exploitation of such advanced concepts as cognition and trust and
OCR for page 16
2007-2008 Assessment of the Army Research Laboratory could measure the performance improvement of radio networks. This then could be used as motivation for further development of this radio system for deployment in wider systems. The use of advanced computing for basic science is still evolving, with several new projects showing promise. A project to develop a code for designing microfluidic devices using large (1600 processor) supercomputers represented excellent science teamed with a deep understanding of how to leverage supercomputing. This project has potentially significant Army applications such as designing instruments to detect biological warfare agents. Another project to understand quantum dot formation is particularly relevant to night-vision sensors and is being pursued through an active collaboration with universities using state-of-the-art multiscale techniques. Another project involves the calculation of binding affinities of protein-ligand complexes and focuses on dealing with ricin and other toxins, again clearly a relevant problem. The project demonstrates good algorithmic techniques. However, questions remain: Why was existing open-source software such as NAMD (NAnoscale Molecular Dynamics) not used?2 Is there a plan for using the Army’s supercomputing systems with such software? Leveraging such widely accepted codes should be used up to the point where deficiencies are present but should accelerate the pace of reaching at least first results. Significant Reorganizations for the Future There have been continued improvements both in the stability of the key CISD management team and in the directorate’s reorganizations that represent changes to support a changing research portfolio. Perhaps the single most important organizational accomplishment since 2006 in terms of the significant dividends that it should produce for the foreseeable future is the creation of the Network Science Division and the related Mobile Network Modeling Institute. According to ARL, this institute is an outgrowth of prior Board assessments which had indicated that a variety of issues associated with mobile networks, especially mobile ad hoc networks (called MANETs) in which network nodes come and go in time, had risen to be of crosscutting importance to ARL. The issues that the Board had discussed included network security, ad hoc wireless networks in particular, system prototyping, and model validation and verification. The structure of the new NSD focuses on three levels of the problem: tactical network assurance, networking sciences development, and the sustaining of base network assurance. This focus represents an excellent capability, if executed well, for addressing the issues of “I can’t define the problem precisely” in networks at all scales, for integrating the efforts of engineers and scientists (especially mathematicians), for starting solid validation and verification efforts for new network technologies and applications that will significantly simplify downstream system deployments resulting from divisional research, and for developing serious transition plans for such technologies. This approach should be a mechanism to allow ARL to track and evaluate in very fundamental ways new technologies that come out of both the academic and commercial worlds (such as the entry of search engine firms into the cellular telephone and cellular telephone applications arena). The first class of problems on which NSD focuses addresses lower-level issues associated with tactical and battlefield wireless networks, from signal processing to intrusion detection, and places under one roof several of the projects that had been scattered across other ARL directorates in the past. Multiple quick-reaction laboratories have been folded into this division, which should aid in meaningful early 2 NAMD is an open-source parallel molecular dynamics code designed for high-performance simulation of large bio-molecular systems. NAMD can simulate the movement of proteins with millions of atoms, making it the world’s fastest parallel molecular dynamics program.
OCR for page 17
2007-2008 Assessment of the Army Research Laboratory evaluation of new networking infrastructure, and perhaps even more important, of how new applications may play on top of such networks. This capability should enhance ARL’s overall laboratory facilities significantly. The second class of problems that NSD focuses on appears to be still in its formative stages but has as its goals a much higher level understanding of networks and networking in general and the management of a variety of collaborative ventures in the networking area. The latter effort includes the existing Communications and Networks CTA and the new Network and Information Sciences ITA. The proposed creation of a new Network Sciences CTA and that of an Army-wide Information Assurance Center of Excellence seem to be appropriate moves to expand ARL’s capabilities in significant ways. The third class of problems on which NSD focuses picks up on continuing issues of security for more classical networks, as was practiced by personnel in the prior High Performance Computing Division for systems across the Army and DoD. What is especially important here is that this new organization inserts some ability to step back from the day-to-day problems of intrusion detection and think a bit more globally and longer term, at a time when more and more MANET-like systems are integrated among traditional networks. The Mobile Network Modeling Institute works with external and internal organizations on end-to-end models of MANETs for tactical purposes before they are developed and to allow those models to guide both the development and deployment activities. These models range from environmental components (these simulate transmission characteristics in battlefield environments under a variety of weather conditions and terrains), to components relating to the signal processing needed in the network nodes for shaping transmissions and receiving them, to the software protocol layers that manage communications, to the applications layers that use the communications paths. It is expected that such models would be hosted on Army supercomputing facilities, with an intent to archive the data produced using structured format (XMDF, or eXtensible Model Data Format) for validation and later use. This institute is clearly appropriate, with the potential to develop large-scale networked radio codes strongly matched to emerging Army needs. However, there is a potential danger of the work of the institute’s being too ambitious and oversold, especially if there is not a strong experimental component to validate and verify the models. The development of high-performance computing codes by themselves is no substitute for good physics to start with, for example in ensuring adequate terrain models that can produce the accuracy and precision needed in the propagation models. Building a connection with other Army sites such as Yuma Proving Ground, Arizona; Fort Irwin, California; Fort Dix, New Jersey; and others may provide valuable sites for such experimental validation in different surface environments. A related concern involves how some of the early projects associated with this institute, discussed in detail later, balance good academics with Army needs, particularly when optimization of some aspect of a project is part of the effort. The Blue Radio, in particular, seems to be an ideal candidate for building and then validating models of sensor networks that can be verified experimentally, but CISD did not evince any explicit recognition of such an opportunity. Also, before large-scale archiving of modeled data begins, serious thought must be given to how to correlate these data with experimental data, and perhaps how future, yet-undefined projects may want to leverage the archived data for fast development of new systems or quick response in order to determine how a deployed system might perform in a new environment. The other major reorganization within CISD was the conversion of the former High Performance Computing Division into the new Advanced Computing and Computational Sciences Division. The revised mission of the division is to advance computational sciences and HPC technologies in support of Army systems. Most of the division, however, continues as in the past to support the computational infrastructure for ARL, particularly the DoD Major Shared Resource Center and the Army High Performance Computing Research Center—both state-of-the-art supercomputing facilities. Only part of
OCR for page 18
2007-2008 Assessment of the Army Research Laboratory one branch, the Computational Sciences and Engineering Branch (CSEB), performs research that falls within the scope of the Board’s assessment. However, the research directions of this group have been broadened, specifically to include a focus on how to get supercomputing levels of computational power into multicore embedded systems that can be positioned closer to where they are needed. In a real sense, this reflects the view that, for the Army, the machine room of the future will be the battlefield and not just a large, air-conditioned center in the United States proper. In particular, the following moves are commendable: considering how to get petaflops in a truck (a petaflop being the level of performance for the fastest machines today); examining how to leverage the emergence of new and specialized computational engines such as the multicore microprocessor chips that are becoming ubiquitous in everything from servers to laptops and PDAs, graphics and game processor chips with extraordinary computational capability that can be used for functions other than graphics, and field-programmable arrays (semiconductor devices that can be configured after manufacturing) for specific Army applications; and beginning to examine the use of such capabilities in basic science projects to explore alternative technologies from the nanoscale or biological realms. A small effort has been started to establish an asymmetric computing center to explore the use of many of these nontraditional computing architectures for specific Army applications. This should be encouraged. OPPORTUNITIES AND CHALLENGES Systems Engineering As noted in ARLTAB assessments in prior years, a significant challenge remains in ensuring that even in relatively basic research programs sufficient consideration is given to questions about how potential systems that might be developed out of such research could be deployed and used in real Army scenarios. The Board continues to suggest that a small amount of systems engineering early in many programs could help avoid paths that, even if successful, would be difficult to deploy in those systems that require real-time responses; conversely, this same effort would provide insight into alternatives that would mesh much better with practice. The same systems engineering focus would also help early on to compare research goals with expected roadmaps for established technologies and would help prepare realistic statements of the potential gains from the new technologies being researched. One example of current research areas where such a focus might be valuable is the deployment of sensor networks, especially for chemical and biological agents where now a single detector is deployed. A corresponding determination of what is needed in terms of additional computational or network support is essential to achieving a viable detection system. One particularly strong project in this arena (on microfluidic sensor design) was being done by university scientists, and while they clearly understood the potential applications to Army missions, there was little or no thought expressed as to how, if the project is successful, transitions to deployable systems might take place. Such systems (or systems of systems) would integrate many functions, including sensing, analysis of responses, and communication of information. The use of polarized thermal imagers is another example in which the desire exists to transition something to Army use, but to do so more quickly it would help to extrapolate potential system requirements from a suite of possible deployed configurations. Such requirements may prove invaluable in specifying appropriate sensor detection characteristics and in developing an understanding of the kinds of outputs that need to be generated by the associated processing system. Similarly, projects that attempt to provide autonomous navigation for small robots in an urban setting where the Global Positioning System may not be available are clearly of significant value to soldiers in
OCR for page 19
2007-2008 Assessment of the Army Research Laboratory terms of reducing operator workload. However, simply trying to use existing computationally expensive image-recognition algorithms for building recognition may not be feasible for small mobile platforms, especially given the potential size of the required three-dimensional image database and accounting for the effects of battle damage on buildings. Some systems engineering to bound the amount of potentially available computational resources and suggest hybrid methods of navigation (using compasses, simple inertial navigation devices, and simpler image programs that construct line segments, vanishing points, and other geometric features) may in fact provide more direct and implementable solution plans. This would be true particularly if and when packs of multiple robots of this type were to be employed, and tasks such as sweep and survey in mass became important. Some of the start-up efforts within the NSD in conjunction with the Mobile Network Modeling Institute may invoke similar concerns. An attempt to develop a science of networks at multiple levels is proceeding in a reasonable fashion mathematically, but it may be more significantly advantaged by the inclusion of a greater focus on whole systems and on overall network performance as seen by the end user in real environments—especially for those cases where the user is a soldier and the environment is a battlefield. This relates to the question of how to balance good academics with Army needs, particularly when attempts are made to optimize without a solid estimate of what the optimization will buy and a clear understanding of whether the fundamentals are well enough known at the current time to justify an optimization study. In particular, it may be beneficial to have some hardware experiments that run side by side with the theoretical work to demonstrate the applicability of the latter. One such collaboration that may pay exceptional dividends is to model sensor networks enabled by the Blue Radio project under way within CISD. Such collaboration might result in somewhat fewer narrowly focused publications, but it would produce results that would be making a more important contribution to the research in mobile ad hoc wireless networks that will be used in the future by the Army. Also, as discussed earlier, some serious system engineering thought up front as to how any archived data resulting from the institute’s modeling or experimental data might be used for future problem solving, and subsequent use of that insight to help organize the archiving effort properly, might provide a very beneficial long-term resource to ARL. Validation and Verification Another general area addressed in prior ARLTAB reports that still remains a challenge is the testing and evaluation of experimentally driven programs and the validation and verification of models—validating that models developed during research programs actually reflect reality and verifying that codes or systems that are supposedly constructed to match are in fact correct implementations of the models. There are research areas such as machine translation in which these activities are central to the research process and others with apparently little such focus. In other research areas, particularly in projects involving complex computations for basic science, there is still a tendency to developing stand-alone codes without any clearly articulated approach to ensure that both the algorithm modeling the physics and the implementation of that algorithm are correct. A tendency to believe the machine is evident and needs to be avoided by formal verification. Further, very often these codes are in areas where the community as a whole does have standard open-source codes, such as NAMD for molecular dynamics, and those codes have already been adapted for execution on supercomputers, with which the Army is well equipped. The use of such codes outright for the computations and for the verification of new codes through careful side-by-side comparisons is warranted. Toward standardizing its practices, CISD should examine the methods by which other government and industry laboratories perform effective and efficient verification and validation.
OCR for page 20
2007-2008 Assessment of the Army Research Laboratory Closely related to this issue is the challenge of obtaining comprehensive data sets to validate models completely. In many areas, such as weather or atmospherics, ARL’s facilities for gathering relevant data are first rate. However, in other newer areas, such as chemical or biological agent detection, complete data sets may not be immediately available at ARL facilities, and alternative mechanisms or partnerships for gaining access to such data should be sought. Some of the newer laboratories such as the Wireless Emulation Laboratory are moving to a largely simulated environment, but without a strong plan to validate, at least periodically, such simulations against the real world, the results emerging from such facilities may not have the grounding in reality to make the results applicable to real Army problems. Another emerging general challenge that is appearing in many programs is an increasing need to perform sophisticated analyses on experimental data. Such analyses involve both classical statistical computation and, perhaps more importantly, information extraction from large and often unstructured data sets. Data mining has emerged in the commercial world as key for applications ranging from determining personalized online purchase preferences to performing portfolio analyses. Similar techniques will become of increasing importance for CISD-relevant applications ranging from sensor network data analysis to prognostication and prediction of the dynamic health of platforms ranging from vehicles to aircraft. These data-mining techniques will also become important in looking through reams of multi-dimensional experimental data to develop and validate new detection algorithms and to analyze massive intelligence data sets for the detection of potential terrorist activities from non-physically-based data. The results from the models emerging from the Mobile Network Modeling Institute are examples of such data sets with potentially long-term value, but for which the schemas used for archiving and then retrieving them will make all the difference later on as to their ultimate usefulness. Growing in-house expertise in such areas should provide ARL with opportunities for both off-line and online system implementations with extraordinary increases in autonomy and robustness. If done properly, such implementations will blend in seamlessly with more traditional numeric-oriented signal processing to produce intelligent and agile real-time control loops for a wide spectrum of future Army systems. A closely related suite of problems of increasing importance to the Army lies in the ability to accumulate, analyze, understand, and efficiently process human and electronic intelligence about relationships between individuals and organizations in an asymmetric battlespace. This capability was referred to as “ information fusion” in the 2005-2006 assessment report. CISD has recognized the importance of networks in general by the establishment of the new NSD, and it is clear that several new projects within ISD are oriented toward getting up to speed specifically on information-fusion applications. In addition, ARL has articulated a goal of developing a laboratory-wide network science research program to address such global issues. However, even the early projects observed during this review cycle before such network sciences programs are fully fleshed out, if continued in relative isolation, may not materially advance ARL’s capabilities, especially given the large number of other organizations pursuing similar activities. One suggestion might be to mirror the careful experimental setup and evaluation work done for the prior CISD machine translation work and to focus significant attention on three aspects of the human intelligence problem: understanding what metrics are most valuable to the Army in the field, obtaining or developing realistic but unclassified data sets (such as from gang databases or local police databases), and developing rigorous validation procedures that determine the potential effectiveness both of individual algorithms and of hybrid approaches (as was done in the machine translation arena). Of particular importance, and where perhaps only ARL has the time frames and overall expertise, are issues of scale—what happens as such databases grow to huge sizes and/or are created as collections of separate and localized databases. In a related vein, increasing joint activities with the ISD, NSD, and Mobile Network Modeling Institute may be of value, since there is an increasing understanding that
OCR for page 21
2007-2008 Assessment of the Army Research Laboratory the properties of time-changing networks of all scales, from local sensor networks, to the connections represented by Internet traffic, to the social networks exhibited by both civil and terrorist groups, all obey similar properties, and expertise in one may give significant insight into another. Also, the potential value to the Mobile Network Modeling Institute of using the Blue Radio as a platform for demonstrating and validating underlying models cannot be underestimated. This is a platform that ARL understands (because ARL designed it) and that was built for an Army application (ground-level sensor networks) that has few if any commercial counterparts (cellular telephones, for example, operate several feet above the ground, with communication with antennas that are quite tall and have different power characteristics from those of the Blue Radio). The development of the Asymmetric Computing Laboratory should open up significant cross-divisional opportunities. For example, one of the concerns regarding the development of lightweight LIDARs is in the associated computation. Investigating the potential to host such applications on alternative execution platforms seems a natural fit. Other Areas of Relevance to the Directorate While the weather modeling efforts within the BED are largely state of the art, challenges exist when trying to move beyond the atmospheric physics to the use of such models for building real applications on top of them. Many of these challenges may be overcome by cross-divisional efforts within ARL. Such collaborations may help, for example, in the use of weather data for the routing of UAVs, for which simple but static algorithms that do not account for weather movements, vehicle dynamics, and formation flying are liable to be inflexible and unverifiable. An understanding of the state of the art in route planning, for both military and commercial aviation, is essential, as is an understanding of how to create more robust, adaptable, and dynamic algorithms. Related projects in decision aids for warfighters that try to use statistics or fuzzy logic to incorporate weather conditions into command decision tools seem to suffer from similar problems of pushing classical algorithms too far. A paradoxical observation apropos of the highly successful engineering of the Blue Radio, discussed earlier, is that this design effort used state-of-the-art, but largely discrete, components to build a prototype that is clearly acceptable for demonstration purposes but is not at the state of the art in terms of implementation as a single chip that would be needed for a deployable system. There is a real need for the Army to take control of the technology, but the problem is a lack of in-house very large scale integrated chip design experience. Digital chip design can be done relatively easily today (at least for the level of complexity exhibited in the Blue Radio), but analog design, especially for the RF links needed to make a single-chip solution, is much more specialized and probably not something that ARL should invest in at this moment. A collaborative effort, perhaps with a CRADA, may be more appropriate. In any case, it may also be of value for ARL to have enough in-house expertise at least to size such chips roughly and then to project how advances in technology may result in improvements over time in system metrics such as size, power, complexity, and heterogeneous integration (e.g., of complementary metal oxide semiconductor cores and analog technologies through emerging three-dimensional technologies). In prior ARLTAB reviews, the Board has commented on both the importance of high-performance computing to ARL’s and the Army’s mission and on the need for sufficient resources to target relevant research with maximal long-term impact. The reorganization of the AC&CSD and a specific focus on high-performance embedded systems should help. There also has been a noticeable improvement in the quality of the research at the AHPCRC. However, significant challenges still exist. Clearly there still is a need for additional research and development resources, specifically for developing a professional staff that is capable of building HPC software products which are efficient and application-specific and
OCR for page 22
2007-2008 Assessment of the Army Research Laboratory for a complementary activity to transition such software into real systems. This resource will become increasingly important as successes are obtained in bringing new supercomputing hardware (e.g., multi-core) into the embedded system space. There is still a lack of an HPC vision in the ARL divisions other than AC&CSD, a lack that will impede the Army’s capability of migrating new applications—for example, advanced weather codes or hybrid language translation systems—into combat systems for direct battlefield use. The question is still open on how to leverage systematically both the embedded and the two supercomputer facilities across all activities, including crossovers into nontraditional computing-intensive applications such as signals intelligence. The Mobile Network Modeling Institute is perhaps the first instance in which the use of HPC resources has not just been part of individual projects but has become a unique enabler that is essential to achieving the institute’s mission. While serious attempts to use emerging computing devices are laudable, as in the establishment of the Asymmetric Computing Laboratory, there is the danger that more efficient alternative algorithmic approaches may be overlooked. Evidence of this is the recent Stanford vehicle that won the Defense Advanced Research Projects Agency’s autonomous vehicle grand challenge by using simple machine learning techniques rather than complex and very computing-intensive, specialized image processing. Crosscutting Issues of Relevance to the Directorate Almost all of the crosscutting issues discussed above are of direct relevance to CISD activities or have aspects that could benefit strongly from CISD involvement. The crosscutting issue of microrobotics, for example, has strong roots in all CISD divisions. Clearly, autonomous control, location-identification and trajectory control, and surveillance sensing are squarely in line with current ISD activities. NSD’s potential involvement ranges from communication with an individual microrobotic vehicle to managing the behavior of a swarm of such vehicles. BED has involvements on two ends—in using data from microrobots (especially micro-UAVs) to support real-time battlefield weather forecasts or to make predictions about chemical, biological, radiological, and nuclear plumes, and as a user of such forecasts in computing flight paths compatible with vehicle capabilities. AC&CSD is moving toward expertise in embedded high-performance computing—the kinds of computing that would be needed both aboard a microrobot and in the command-and-control links needed to link it into the battlefield. In addition, the ability to do high-fidelity design, simulating, and modeling of the microrobotic platforms during the design phase and in mission planning and rehearsal would be of significant value. In the next major crosscutting issue, power, CISD clearly must be a key player, both in using information processing to help optimize the power used by a platform overall to perform its mission and in developing information processing systems that are energy-efficient in their own right. Similarly, in the areas of prognostics and diagnostics CISD needs to be involved both in platform-based fault detection and reconfiguration and in remote real-time data mining, parameter extraction, trend analysis, and real-time modeling. While CISD does not have as central a role in biomechanics as that of the Human Research and Engineering Directorate, there certainly will be a need to develop and then support significant modeling activities, particularly using HPC expertise, facilities, and resources. Acoustics is an area that has already been mentioned as a strong point in CISD’s research portfolio. This importance will increase as additional sensors and additional laboratories such as HRED’s Environment for Auditory Research (EAR) come online and require modeling support, data visualization, and correlation with atmospheric effects.
OCR for page 23
2007-2008 Assessment of the Army Research Laboratory Modeling and computational sciences clearly overlap multiple components of CISD’s charter and have been an area identified for crosscutting activities in previous Board assessments. They remain so. The issue of identifying potentially disruptive technologies that might radically change the problems confronting the Army (such as the rise of asymmetrical warfare and IEDs) and the way that the Army needs to leverage technology to respond to them are of laboratory-wide importance. However, in the very fast-moving areas that are the realm of CISD, technology changes, representing both threats and opportunities, occur faster than in most other areas. Therefore, each of CISD’s divisions, and CISD as a whole, may benefit from an explicit recognition of the potential of such technologies and the development of a formal mechanism to help identify them in a timely fashion. OVERALL TECHNICAL QUALITY OF THE WORK One of the assessment criteria applied by the Board asks if the scientific quality of a directorate’s research is of comparable technical quality to that executed in leading federal, university, and/or industrial laboratories both nationally and internationally. As in prior years, the answer to this question is that it is generally true for the Computational and Information Sciences Directorate, with exceptional expertise in selected areas such as weather, intelligent optics, and machine translation of foreign languages. Some particularly strong projects reviewed in this cycle were, however, staffed by university scientists, not ARL personnel, and it was unclear how much of the research had been done by or transferred to ARL, and to CISD in particular. Other areas, such as networking sciences, where formal expertise was recognized as needing growth, have seen significant additional resources and organizational changes made to improve them. However, there are still some key areas, such as the capability to develop and deploy HPC-based applications and multicore programming, where additional and improved expertise would be broadly beneficial. While the CISD’s scientific and engineering staff are, on the whole, conducting and publishing quality research in a number of areas, there does not seem to be much involvement in leading scientific societies and organizations or sufficient attendance at top-tier research conferences. Promoting such involvement should give rise to more scientific recognition and stature for the research staff, make them more aware of the state of the art in other groups, and make the laboratory as a whole more attractive to new Ph.D.’s. Sufficient funding should be provided to ARL so that funding is not a constraint on managers’ ability to enable the interactions of ARL staff with the scientific community through travel to professional meetings. A second criterion applied by the Board asks if the research program reflects a broad understanding of the underlying science and research conducted elsewhere. The answer here is mixed; the areas mentioned above as being exceptional are also the areas in which there is a good understanding of the state of the art elsewhere. This is especially true for areas that have emphasized testing and evaluation, such as machine-based language translation. However, in other areas such as route planning, use of field-programmable gate arrays, programming global positioning units, and open-source software packages that do not have a history of prior internal projects or collaborations in that area with others outside ARL, there is a distinct drop-off in an understanding of other work or of the availability of existing program packages. A related criterion addresses the qualifications of the research team vis-à-vis the research challenges. With just a few exceptions, the match seems to be present. In addition, the aggressive effort to hire new Ph.D.’s and to encourage Ph.D.-level work by current employees is a very positive indication. Evidence of this is a series of talks over the past 2 years by several Ph.D. students.
OCR for page 24
2007-2008 Assessment of the Army Research Laboratory The next criterion deals with the structure of programs in terms of employing the appropriate mix of theory, computation, and experimentation. The results here are again mixed. In cases where projects take advantage of ARL’s outstanding test facilities and weave in a feedback path that validates theory and drives more robust algorithm and system development, the results are usually strong, with obvious opportunities for transition. In other cases, where one or more of these features are lacking or a bit weak, the efforts might be less than optimal. Examples include using simplistic or older algorithms for route planning and polarimetric imaging, and modeling the operation of polarized light sensors in non-ideal conditions and misalignment. In terms of current and projected equipment, facilities, and human resources, CISD continues to have an appropriate mix to achieve success. This is particularly true of the White Sands, New Mexico, facilities for BED; the intelligent optics laboratories of ISD; the mobile networking laboratories and institute recently established for NSD; and the supercomputing facilities. The development of new facilities such as the Wireless Emulation Laboratory and the Asymmetric Computing Laboratory indicates that ARL is serious about being agile in the face of new technologies. One exception is the need for stronger explicit support for computational scientists and professional HPC staff within AC&CSD for research into high-performance algorithm and program development. The support needed for computational scientists and professional HPC staff should be comparable to the relatively strong support currently provided for infrastructure (e.g., running and managing jobs in machine rooms). Some of the other directorates, such as the Sensors and Electron Devices Directorate, have leveraged the building of new facilities into an attraction for prospective employees. These new facilities within CISD should offer similar attractiveness and should also be used in that way. In some divisions with exceptionally strong experimental and field work, such as BED, it may be worth considering whether or not increasing the number of technicians might free more of the research staff time for those research issues of most importance to ARL. The answer to the question of whether the various research teams are responsive to the Board’s recommendations is a resounding yes. There have been identifiable organizational changes, especially in the past 2 years, that seem to be directly focused on alleviating problems on which the Board had commented in the past. The NSD and associated institutes and other initiatives are a premier example, organizing around an end-to-end focus on networking in the large. The reorganization of AC&CSD to address the growing appearance of HPC-like functionality in everyday battlefield computing resources is another example. Further, within the portfolio of research projects there have been significant changes, with drops in areas that the Board had suggested were redundant or behind the state of the art (such as nanoelectronic devices) and the introduction of new projects in areas where the Board suggested that there was significant Army mission-relevant potential (such as embedded HPC, networking problems, and bio-inspired applications). This responsiveness has even shown up in the way that individual divisions, especially BED and ISD, report out their research portfolios at the assessment reviews. There is, however, still room for improvement, especially in articulating both divisional and overall CISD strategic plans and the rationale behind how the research portfolio is adapted to customer pressures while still maintaining a solid and relevant basic science capability. CISD has shown improvement, especially in BED and ISD, but the improvement is not consistent across divisions and does not articulate as crisply as it could. An emphasis on the core long-term relevant scientific problems and an articulation of short- versus long-term strategic goals would help in continuing to maximize the value of CISD’s research portfolio to the Army. A suggested additional metric might relate to how CISD’s customers perceive the value of their collaborations, with a related discussion of how expectations and requirements are developed in light of such a metric.
OCR for page 25
2007-2008 Assessment of the Army Research Laboratory In addition to addressing the assessment criteria, there are several other observations in a variety of areas. First, while there seems to be a significant number of collaborations of various sorts, it is often not clear how those collaborations really interact with ARL programs (versus simply being funded grants), and what part of the results reported from the collaborations are due to ARL versus external researchers and contractors. This information is important when trying to judge the overall level of expertise of the ARL staff. The proliferation of CTAs and ITAs in particular represents collaborations that have not had as much review as other activities have had, and the Board cannot properly judge their overall effect on ARL’s portfolio. Second, judging the understanding of the state of the art would be aided by more explicit discussion in reviews about CISD’s view of the state of the art elsewhere and by knowing in what metrics one would see improvements as a reflection of success in ARL projects. The work at CISD continues to be generally well targeted on Army needs. The machine translation work continues to drive deployments into the field and helps in the processing of newly discovered document troves. BED continues to keep its Army and national science niche in defining and predicting the characteristics of meteorological phenomena that are critically important to fixing the properties of the atmosphere on time and space scales relevant to rural and, of increasing importance, urban battlefield situations. The growth in focus on networking at multiple levels correlates directly with the growth in the network-centric battlefield and the need to integrate disparate information sources in real time to support decision making. Prior ARLTAB assessments have noted the recognized exceptional contributions of the machine translation work. This continues to be the case. Judging the contributions of much of the rest of CISD to the broader community remains more difficult. There seems to be a significant variance across the divisions in the number of publications, the quality of the publication forums, and the impact of the work. A variety of indexes are used in academia for such purposes; they include the H-index for references and impact factors for publication venues. Data sources for computing such indexes can be found at Web sites such as those for Googlescholar, ISI Web of Science, Science Citation Index, and Citeseer. Performing such self-evaluations in advance of reviews would help both the Board and ARL to identify where the lead contributions are coming from and which venues should be targeted to maximize the exposure of research results.