Past Seminars

If you would like to receive e-mail notification of these seminars please request to be added to the seminars mailing list. A search and subscription feature will be available soon.

Abstracts of Previous Seminars

Abstract:
Integrated Systems Health Management includes such functions as diagnostics, prognostics, condition-based maintenance, reconfigurable control, and mission autonomy. These technology directions are rapidly evolving in aerospace area driven by strong NASA, DoD, and commercial demands. One driver is reducing operations (maintenance) cost, another is enabling autonomy through automated contingency management. Similar requirements drive current related work in many other industries: automotive (telematics), process control, semiconductors (e-diagnostics), information systems (autonomic computing).

This talk will focus on diagnostic estimation and anomaly detection in the context of parametric (as opposed to binary) variables of systems health state. The tutorial part of the talk will present a perspective from SPC to LS estimation to new approaches.

The technical novelty of our work is related to using convex optimization embedded in the decision loop. This allows effective handling of monotonicity constraints (accumulating damage never goes away by itself). Solving a QP problem at each estimation step provides a scalable and efficient
technology in many respects more desirable than particle
filtering.

It also turns out that the considered formulation leads to a very good description of shock and wear processes accurately estimating jumps in decision variables. This is related to recently provded sparsity properties of LP solutions of underdetermined linear equations.

The presentation is illustrated mostly by aerospace applications validated through high-fidelity simulations and real deployments. Estimation of health state in jet engines, rocket launch vehicle, and an uninhabited air vehicle is considered.

Speaker's Bio:
Dimitry Gorinevsky is a Consulting Professor of Electrical Engineering at Stanford University. He received a Ph.D. from Moscow (Lomonosov) University and a M.Sc. from the Moscow Institute of Physics and Technology. He held research, engineering, and academic positions in Moscow, Russia; Munich, Germany; Toronto and Vancouver, Canada. He spent 10 years with Honeywell, for last six years representing Honeywell Labs in Silicon Valley. His interests are in decision and control systems applications across many industries. He has authored a book, more than 130 reviewed technical papers and a dozen patents. Dr. Gorinevsky is an Associate Editorof IEEE Transactions on Control Systems Technology. He is a recipient of Control Systems Technology Award, 2002, and Transactions on Control Systems Technology Outstanding Paper Award, 2004, of the IEEE Control Systems Society. He has been elected a Fellow of IEEEfrom Control Systems Society, class of 2006.

Affiliation(s): Graduate School of Information Science and Technology at Tokyo University and Graduate School of Business at Tsukuba University, Tokyo, Japan

Abstract:
A "chance" is an event or a situation significant for making a decision in a complex environment. Interdisciplinary discussions by researchers from philosophy, sociology, artificial intelligence, finance, complex systems, medical science, etc, have contributed to

the methodologies of "chance discovery" since it was initiated in 2000. More recently, companies responded to this research and came to organize the Chance Discovery Consortium in Japan. In this talk, methods and applications of chance discovery are presented. Visual data mining methods developed for chance discovery aid the thoughts of individuals and a groups about valuable scenarios in the future. This discovery process is called Double Helix, where humans and computers cooperatively make spirals of deepening attention to chances. In this process, highly valued and feasible future scenarios in the environment emerge, like creatures emerging with new chromosomes. The emerging scenarios motivate the user to work in the real world to try actions, and the new data acquired from these actions accelerates the process. Chance discovery, in other words, is the child, and is also the parent of scenario emergence.

Speaker's Bio:
Prof. Ohsawa coined the term "chance discovery", and initiated a series of international conference sessions and workshops that defined the field. Since 2003, he has been the general chair of Chance Discovery Consortium Japan, linking researchers in cognitive science, information sciences, and business sciences to Chance Discovery. He is collaborating internationally to develop and apply chance discovery methods, for example, with the DISCUS Project (Distributed Innovation and Scalable Collaboration in Uncertain Settings) at the University of Illinois at Urban Champaign; the Project Based Learning Laboratory (PBL) of the Center for Integrated Facility Engineering at Stanford University; and the Congnitive Science Department of Kingston University. He is the author of Author of Chance Discovery , Springer Verlag.

About the book:
Chance discovery means discovering chances - the breaking points in systems, the marketing windows in business, etc. It involves determining the significance of some piece of information about an event and then using this new knowledge in decision making. The techniques developed combine data mining methods for finding rare but important events with knowledge management, groupware, and social psychology. The reader will find many applications, such as finding information on the Internet, recognizing changes in customer behavior, detecting the first signs of an imminent earthquake, etc. This first book dedicated to chance discovery covers the state of the art in the theory and methods and examines typical scenarios, and it thus appeals to researchers working on new techniques and algorithms and also to professionals dealing with real-world applications.

Abstract:
Recently I have been working in the area of Educational Programming that is defined to be the intersection between education and programming. I propose that the results in this talk should be part of every programming language curriculum.
For the purposes of this seminar, define program Integrity to mean firm adherence to a code of engineering, scientific, and aesthetic values in programs. In this seminar I use program schemas (explained below) to precisely state and in some cases prove some control structure program integrity results. The use of schemas enables us to avoid what Alan Perlis called the “Turing machine tar pit,” i.e., every class of programs is more or less equivalent to every other class of programs.

A program schema is defined to be a program with uninterpreted primitive procedures. Program schemas provide an important tool for precisely stating and proving some control structure integrity properties. For example, given two classes of program schemas, X and Y, X will be defined to be strongly more powerful than Y (written X _ Y) if and only if the following two conditions hold:

For every program schema g in Y, there is an equivalent program schema f in X such that f is just as efficient in space and time as g.
There is a program schema f in X for which there is no equivalent program schema in Y.

We present (many of these are not new cf. [Paterson and Hewitt], etc.) the following results for classes of program schemas:

linear recursiveclass of programs that are recursive programs such that each control path trough the program has at most one recursive call which is top level and at the end. This class is strongly equivalent to nonrecursive programs using assignment and goto.
sequentialrecursiveclass of programs that are recursive programs with left to right sequential evaluation of arguments to procedures.
parallel recursive class of programs that are recursive programs augmented with parallelOr, parellelAnd and futures [Baker and Hewitt]
concurrentclass of programs are recursive programs augmented with the ability to dynamically create Actors with locally updateable state + asynchronous messagingsuch as Java, C#, Actor languages, etc.

Each of the above represents a fundamental program archetype. It is important to characterize the computations that can be carried out by each of these archetypes.

Universality is a control structure integrity issue that can be addressed as follows: define a set S of program schemas to be strongly universal if and only if for every computation C that is carried out according to the laws of physics, there is a schema g in S such that g carries out C and g is just as efficient as C in space and time. In this way, a generalization of Church’s thesis can be formulated as the following conjecture: Concurrent program schemas that are augmented with the usual primitives for arithmetic, string, array, quantum computing, and coin flipping are strongly universal. It will be interesting to see if some development, e.g. in nanotechnology, molecular computing, etc., proves this conjecture wrong.

Speaker's Bio:
Carl E. Hewitt is an emeritus professor at MIT. In his dissertation, he introduced Planner, a new kind of programming language based on using Plans to process Goals and Assertions. With his students, he developed the Actor model of computation based on Actors that asynchronously send messages that are processed by (other) Actors to dynamically change their own local state, create more Actors and send more messages. More recently he has worked on the development of Just in Time reasoning systems and Educational Programming (which is defined to be the intersection of education and programming).

Title: "A Unified Approach for the Integration of Distributed Heterogeneous Software Components"

Abstract:
A framework is proposed for assembling software systems from distributed heterogeneous components. For the successful deployment of such a software system, it is necessary that its realization not only meets the functional requirements but also non-functional requirements such as Quality of Service (QoS) criteria. The approach described is based on the notions of a meta-component model called the Unified Meta Model (UMM), a generative domain model, and specification of appropriate QoS parameters. A formal specification based on Two-Level Grammar is used to represent these notions in a tightly integrated way so that QoS becomes a part of the generative domain model. A simple case study is described in the context of this framework.

Speaker's Bio:
Barrett Bryant received his B. S. degree from the University of Arkansas at Little Rock in 1979, and the M. S. and Ph. D. degrees from Northwestern University in 1980 and 1983, respectively, all in computer science. He is currently a Professor and Associate Chair of Computer and Information Sciences at the University of Alabama at Birmingham. His research interests are the theory and implementation of programming languages and object-oriented technology. He has held visiting research positions at the IBM Palo Alto Scientific Center, the IBM Data Base Technology Institute, the Air Force Rome Laboratory, the Army Research Laboratory, Oak Ridge National Laboratory, NASA (National Aeronautics and Space Administration) Marshall Space Flight Center, and the Naval Postgraduate School.

Title: "Decision and Uncertaintly Management for Human and Human/Agent Teams"

Affiliation(s): Emeritus Professor of Mechanical Engineering Design at Oregon State University

Abstract:
In order to better support designers and managers Dr. Ullman has studied and modeled the product evolution process since the mid 1980s. He has performed detailed cognitive studies on individuals and teams, and built systems to capture and query process information. As he learned more, he became increasingly focused on the management of decisions made during technology and business development.

In the late 1990s he led an effort to develop Bayesian methods to help support distributed teams making decisions when their information is uncertain, incomplete, conflicting and evolving. These tools have been refined and applied in industry to support business and technical project teams designing products, making portfolio decisions and selecting vendors. Results show significant increase decision confidence, improved efficiency, better transparency and objectivity, and rationale capture.

Ullman is beginning to extend decision management techniques to operational human-agent teams. In this situation, teams of humans and software agents are working in real time to manage a system. The software agents may be part of satellites, rovers or other devices far removed in time and location from the human team. These agents are gathering information and making autonomously decisions on their own. However, as change occurs, and uncertainty increases or as future scenarios need to be forecast, an agent may need the involvement of other agents and human partners as part of a decision making team. Certain key decisions made by this team may determine the success of a program or the survival of an agent or astronaut.

This presentation will develop the importance of decision and uncertainty management in technical teams and introduce the tools developed to support members distributed in time and location. Ullman will outline how these tools are being extended to support human/agent teams making decisions in complex systems.

Speaker's Bio:
Dr. Dr. David G. Ullman is an Emeritus Professor of Mechanical Engineering Design at Oregon State University and is the President of Robust Decisions, Inc. where he developed software tools to support decision management. He is a graduate of the University of Cincinatti and earned his PhD from Ohio State University. Dr. Ullman has authored over 30 technical papers, three books including "The Mechanical Design Process" (3rd edition) and "12 Steps to Robust Decisions", and holds several patents.

Speaker(s): Dr. Luciano Serafini (joint work with Paolo Bouquet and Stefano Zanobini)

Abstract:
Semantic coordination, namely the problem of finding an agreement on the meaning of heterogeneous semantic models, is one of the key issues in the development of information integration systems and semantic web applications. In this talk we propose a new algorithm called CTXMATCH for discovering semantic mappings across hierarchical classifications based on a new approach to semantic coordination. This approach shifts the problem of semantic coordination from the problem of computing structural similarities (what most other proposed approaches do) to the problem of deducing relations between sets of logical formulae that represent the meaning of concepts belonging to different models (here, classifications). We show why this is a significant improvement on previous approaches, and present the results of preliminary tests on two types of hierarchical classifications, namely web directories and catalogs. CTXMATCH has been integrated in the distributed knowledge management platform KEEX (by Distributed Thinking). At the end of the talk we will give a practical demonstration of both CTXMATCH and KEEX.

Speaker's Bio:
Luciano Serafini was born in Trento, Italy in 1964. In 1988 he received his Master's Degree in Computer Science from the Computer Science Department of the University of Milan, with a thesis in the area of logics for knowledge representation and reasoning in multi-agent systems. Since 1989 he has worked at ITC-IRST (Centro Per la Ricerca Scientifica e Tecnologica) in the Automated Reasoning Area. He has done research in the formal development of logics contexts and in the application of contexts in the areas of multi-agent reasoning (belief, desire and intention) and federated databases. He has participated in application projects about Artificial Intelligence Integrated Systems, information navigation, information retrieval, and knowledge management. He has coordinated basic research projects about logics for knowledge representation and reasoning. Dr. Serafini has published papers in the major international journals (among them: Journal of Artificial Intelligence and Journal of Logic and Computation) and conferences such as the IJCAI and ECAI.

Affiliation(s): Assistant Professor in the G.W. Woodruff School of Mechanical Engineering at Georgia Tech

Abstract:
Design is an information-driven process. Starting from information about requirements, designers gradually develop a complete, detailed description of a system that meets these requirements. In this process, a variety of information and knowledge is applied: working principles, behavioral models, detailed models from previous product generations, etc. With the rapid development of networked computing and information technology capabilities, we have an opportunity to improve the support for this information-driven design process significantly.

In this presentation, I will first present an abstract perspective of how information and knowledge is used within the product life-cycle. Based on that abstract model, I will then high-light two topics in more detail:

Speaker's Bio:
Dr. Paredis has a broad, multidisciplinary background. In his research, he combines aspects of information technology, modeling and simulation, and systems theory to support the design of mechatronic systems. The goal of his current research is to develop an integrated IT framework for simulation-based design. He is also still active in the robotics and mechatronics areas in which he did his Ph.D. research. Dr. Paredis is an Assistant Professor in the G.W. Woodruff School of Mechanical Engineering at Georgia Tech. He received his M.S. degree in Mechanical Engineering from the Catholic University of Leuven (Belgium) in 1988, and his M.S. and Ph.D. in Electrical and Computer Engineering from Carnegie Mellon University in 1990 and 1996, respectively. From 1996 to 2002, he was a Research Scientist in the Institute for Complex Engineered Systems at Carnegie Mellon University.

Abstract:
Digital libraries typically serve a specialized community of experts sharing a relatively narrow focus, such as some aspect of law, science, technology or business. Furthermore, users of such libraries tend to have sophisticated information requirements which are not well addressed by traditional search methods. We present preliminary work with a search method that employs a knowledge base tailored to a library's content to improve the accuracy, depth and usability of the information retrieved. By using a knowledge base which contains general background information on botany and climate change research, we will show how such a set of logical axioms and rules can make informed connections between documents and queries. Our system tries to ferret out the documents providing information implicit in a user's query, even if it is not directly specified. For example, if a user's query involves a process such as photosynthesis, the system itself supplies information on how photosynthesis works, what it accomplishes, the chemicals it uses, and so on. If documents within the digital library touch on any of these things, they are potentially relevant. Furthermore, the system can show why it is showing these documents to the user, based on the logical connections within its knowledge base.

Speaker's Bio:
Mr. Crick is a researcher and engineer at GCAS, Inc., in San Marcos CA. He earned his MS at Harvard University, conducting research in probabilistic reasoning and computational linguistics. In addition to his NASA-funded work on search methods, he is also currently engaged in other projects related to reasoning and decisionmaking under uncertainty.

Title: "MAPGEN: A Tale of infusing a mission-critical application for the Mars Exploration Rovers (MER) project"

Affiliation(s): NASA Ames Research Center/RIACS, Autonomy and Robotics Area, Computational Sciences Division

Abstract:
In September 2001, the MER mission chose to use MAPGEN, a ground based decision-support system to help command the two rovers. This talk is the story of the ups and downs in the infusion process what it takes to get to deploy a mission critical application (AI based or otherwise) on a NASA mission.

The Mixed-Initiative Activity Plan GENerator (MAPGEN) is a combination of two legacy systems, one of which was the Planner/Scheduler that flew onboard DS1. The system combines a rich formalism of a flexible temporal constraint network with a familiar front end used by mission operations personnel at JPL to command the two MER rovers. MAPGEN continues to be a part of the uplink command cycle for the Mars Exploration Rovers (see http://marsrovers.jpl.nasa.gov & http://ic.arc.nasa.gov/story).

Mission operators with the help of science personnel, use MAPGEN on the ground, twice daily to build a complex conflict free plan that is packaged and uplinked to command Spirit and Opportunity on the surface of Mars. This generative planner, automatically enforces mission and flight rules encased in a declarative model, as well as constraints imposed to encode the scientific intent of the observations for that Sol, requiring to being done on the surface of the Red Planet.

I will briefly discuss the results of the deployment so far (which have far exceeded our own expectations) and will take you along on the wild ride we went thru to get to be a critical part of NASA's most complex scientific mission to date.

Speaker's Bio:
Kanna Rajan is a Senior Research Scientist and a member of the management team of the the Autonomy and Robotics Area at NASA Ames Research Center Moffett Field, California. He is one of the principals of the Remote Agent Experiment (RAX) which designed, built, tested and flew the first AI based closed-loop control system on a spacecraft. The RA was the co-winner of NASA's 1999 Software of the Year, the agency's highest technical award (http://ic.arc.nasa.gov/projects/remote-agent/).

His interests are in Planning/Scheduling, modeling and representation for real world planners and agent architectures for Distributed Control applications. Prior to joining Ames, he was in the doctoral program at the Courant Institute of Math Sciences at New York Univ. Prior to that he was at the Knowledge Systems group at American Airlines, helping build a Maintenance Routing scheduler (MOCA) which continues to be used by the airline 365 days of the year. He is currently the principal investigator for the MAPGEN Planning and Scheduling software for optimizing science and engineering activity plans for the Mars Exploration Rovers (MER) project.

Kanna is the recipient of the 2002 NASA Public Service Medal and the First NASA Ames Information Directorate Infusion Award also in 2002. He is the Co-chair of the 2005 Intnl. Conference on Automated Planning and Scheduling (ICAPS), to be held in Monterey California (http://icaps05.icaps-conference.org/) and the chair of the Executive Board of the International Workshop on Planning and Scheduling for Space (http://www.congrex.nl/04c06/).

Title: "Diagnosis based on Hybrid Uncertain Models -- Application to Autonomous Satellites"

Abstract:
Modern model-based supervisors use hybrid models where discrete and continuous dynamics interact, to represent complex systems with embedded controllers. Here we present an extension to Livingstone, that uses hybrid models with uncertainty in the form of numerical intervals. We show how autonomous discrete switches slice the continuous state-space into a finite set of regions of behavior we refer to as the configurations of the hybrid system. Our aim is then to identify the set of potential configurations of the system given the commands and observations that are received at each time step. This is achieved by making a prediction of the hybrid state consistent with the grid of configurations. A qualitative version of the configuration is built that allows to interleave a progressive search in the hybrid space with consistency-checks. Results on the simulated attitude control loop of a nominal/faulty earth-orbiting satellite are presented.

Speaker's Bio:
Emmanuel Benazera graduated in Applied Mathematics and Intelligent Systems from the University Paris-Dauphine in 1999, and obtained his Ph.D. in Computer Science from the University Paul-Sabatier of Toulouse, France, in 2003. His research interests are model-based diagnosis, hybrid systems, reconfiguration, filtering, and planning.

Affiliation(s): Biomedical Engineering Center, "Politehnica" University of Bucharest, Romania

Abstract:
The almost complete sequencing of several genomes, including the human genome, as well as the public access to most of this vast amount of information, offer the opportunity to data mine and to explore in depth this unique empirical data depository. In all genomic data bases, genomic information is represented by sequences of nucleotide symbols in the strands of DNA and RNA molecules.

Converting the genomic sequences into digital genomic signals offers the possibility to use signal processing methods for handling and analyzing genomic information. Interesting and, sometimes, surprising large scale features of chromosomes, maintained over distances of 106 - 108 base pairs, up to entire chromosomes, have been revealed by using this approach for both prokaryote and eukaryote.

On the other hand, in the context of analyzing large volumes of data and of presenting the results in a easy to read form, the problem of data representability becomes important. A quantitative approach to data graphical representability is presented and applied to several typical cases of standard signals and for genomic signals. It is shown that the variation of genomic data along nucleotide sequences, specifically the cumulated and unwrapped phases, can be adequately represented as simple graphic lines for low and large scales, while for medium scales (thousands to tens of thousands of base pairs) statistical descriptions must be used.

Speaker's Bio:
Paul Cristea graduated the Faculty of Electronics and Telecommunications (UPB - University "Politehnica" of Bucharest) in 1962, the Faculty of Physics (University of Bucharest) in 1969, and obtained the Ph.D. in Technical Physics (UPB) in 1970. His research and teaching activities have been in the fields of Genomic Signals, Digital Signal and Image Processing, Connectionist and Evolutionary Systems, Intelligent e-Learning Environments, Computerized Medical Equipment and Special Electrical Batteries. He is the author or co-author of more then 125 published papers, 13 patents, and contributed to more than 20 books in these fields. Currently, he is professor and coordinator of the Electrical Engineering and Computers Division of the Faculty of Engineering of UPB, general manager of the Bio-Medical Engineering Center of UPB and director of the Romanian Bioinformatics Society.

Title: "Applying Local Search & Constraint Programming in MER Staff Rostering Program
Rostering Problem"

Abstract:
In recent years there has been an increased recognition of the importance of personnel rostering problems to a wide range of organizations. Personnel rostering can be defined to be the problem of placing resources, subject to constraints, into slots in a pattern, where the pattern denotes a set of legal shifts defined in terms of work to be done. This talk concerns a highly important, large and complex scheduling problem: Staff Scheduling for Mars Exploration Rover Mission 2003 (MER).

The MER scheduling problem consists in generating a configuration of individual schedules which are patterns of days off and on for each staff, over a given planning horizon, positions and rovers. The configuration of staff schedules is generated so as to fulfill collective requirements and the staffing demand coverage while maximizing staff preferences and improving overall productivity.

Speaker's Bio:
Sevugan Alagappan is responsible in developing strategic and operational technical plans that are tied to FriarTuck's business development strategy and research activities. He has expertise in mathematical programming and heuristic techniques.

He was invited by the Asian Productivity Organization to participate as a Resource Speaker in a symposium titled "Technological Innovations in SMEs" held in Taipei, Taiwan. He was one of the lead organizers for Stanford Global Entrepreneurs Challenge, a global business plan competition jointly organized by Stanford University and National University of Singapore. He currently sits in the Judging Committee of Startup@Singapore 2002/3 National Business Plan Competition.

He conducted his industrial attachment project with Andersen, where he worked on security mechanisms and database management of the General Multi-purpose Card Project (GMPC). He won an innovation award for his Research Project by conducting a comprehensive research in solving complex nurse rostering problems in a local hospital. Sevugan completed his
Bachelors (Hons) degree in Computer Engineering at the National University of Singapore.

Abstract:
In this talk I will present the KeY tool, currently being developed at the University of Karlsruhe, Germany and Chalmers University, Gothenburg, Sweden.This tool provides facilities for formal specification and verification of programs within a commercial platform for UML based software development. Using the KeY Tool, formal methods and object-oriented development techniques are applied in an integrated manner. Formal specification is performed using the Object Constraint Language (OCL), which is part of the UML standard. KeY provides support for the authoring and formal analysis of OCL constraints. The target language of KeY based development is Java Card, a proper subset of Java for smart card applications and embedded systems. KeY uses a dynamic logic for Java Card to express proof obligations, and provides a state-of-the-art theorem prover for interactive and automated verification. Apart from its integration into UML based software development, a characteristic feature of KeY is that formal specification and verification can be introduced incrementally.

Abstract:
The Entomopter was a biologically-inspired "micro" air vehicle that leverages the Hawk Moth (Manduca sexta) and the Vampire Bat (Desmodus rotundus) for locomotion and navigation, however the Entomopter is NOT biomimetic... rather, it goes beyond biological mimicry to use techniques not found in the animal kingdom to create a manufacturable vehicle with unique qualities. In particular, the Entomopter is geared for flight and crawling, but uses a chemical muscle that is anaerobic and ignitionless, making is amenable to operation in explosive environments, or anoxic atmospheres such as that of Mars. The wing flapping mechanism takes advantage of resonance (as do the insects), but goes further to create a vibrationless sensor platform that has the potential of slow flight even in rarefied low Reynolds number flight conditions. Unlike nearly all other attempts at flapping wing flight which result in synthesis of bird wings with airplane tials for control, the Entompter can generate not only lift and thrust from its wings, but also full dynamic flight control through the use of circulation controlled unsteady aerodynamics in an autonomically-beating constant frequency wing cycle. This seminar will discuss Entomopter morphology and operation in both terrestrial and Mars environments.

Speaker's Bio:
Robert Michelson is a Principal Research Engineer in GTRI's Aerospace Transportation and Advanced Systems Laboratory. He is also an Adjunct Associate Professor in the Georgia Tech School of Aerospace Engineering. His professional research interests include biologically-inspired robotics; unmanned aerial vehicles; autonomous systems; special purpose digital machines; process control; transportation systems; digital/hybrid computer design; machine vision; physical security/contraband interdiction/anti-terrorism.

Affiliation(s): Computer Science and Engineering Department, University of Washington

Abstract:
Integration of data from multiple sources is one of the longest standing problems facing the AI and Database research communities. In addition to being a problem in large enterprises and collaborative science projects, research on this topic has been fueled by the promise of integrating data on the WWW. In the past few years, we have made very significant progress on data integration, from the conceptual and algorithmic aspects, to the systems and product aspects. This talk will briefly review our successes in data integration, and will describe some significant current challenges. In particular I will discuss the problem of trying to semi-automatically find a semantic mapping between a pair of schemas/ontologies, and describe our work on Peer-data management systems, a flexible architecture for ad-hoc sharing of structured data.

Speaker's Bio :
Halevy's research interests are both in database systems and artificial intelligence. He has worked on many aspects of data integration, query optimization, XML query processing and knowledge representation. He received his Ph.D in Computer Science from Stanford University in 1993, spent 4 years as a member of the technical staff at AT&T Bell Labs, and joined the faculty of the University of Washington in 1998.

Abstract:
Developing correct multithreaded programs is extremely challenging, due to the potential for unintended and nondeterministic interference between threads. A crucial concept for controlling this interference is the notion of atomicity: a function is considered atomic if its behavior is not influenced by the actions of concurrent threads. Atomicity is often achieved using mutual-exclusion locks, and atomic functions are much easier to reason about, since the effect of nondeterministically interleaved actions of other threads can be safely ignored.

In practice, many functions in multithreaded programs are designed to be atomic and their documentation claims that they are atomic. However, because of the difficult of correctly reasoning about multithreaded programs, these claimed atomicity properties are unreliable and often incorrect.

This talk presents the Java Atomicity Checker, a static analysis tool that focuses on specifying and verifying the atomicity of methods and classes in multithreaded Java programs. This tool has caught a number of subtle timing-dependent defects, even in widely used and well-tested library classes such as java.lang.String and java.lang.StringBuffer. These previously unknown defects may cause programs to crash in a nondeterministic and irreproducible manner.

Speaker's Bio:
Cormac Flanagan is a Research Scientist at HP's Systems Research Center. His research interests include programming languages and software engineering tools and technologies. He has worked on a variety of static checking tools.

Title: "Mathematical and Computational Models in Robotics and Structural Biology"

Affiliation(s): Professor of Mechanical Engineering at Johns Hopkins University

Abstract:
Many similar issues exist in the computational modeling of chain-like molecules and highly articulated robot arms. This talk reviews methods of noncommutative harmonic analysis developed by the speaker for computing the entropy of polymers and the propagation of error in robotic manipulators. It is shown how Fourier analysis on the group of rigid-body motions as well as minimization of functions on this group can be useful tools in these calculations, as well as in image registration and encoding of spherical motion.

In the second part of the talk, kinematic models of large biomolecules are discussed. In many biomolecular systems, function is related to structure through motion and conformational change. Methods for coarse-grained mechanics-based analysis of biological macromolecules is a reasonable approach to developing insights into the structural motions that are relevant to function. In this part of the presentation, motions of a number of large biological macromolecules that have been modeled by the presenter, his students, and collaborators are reviewed. These include the maturation of the HK97 virus, normal mode analysis of the GROEL/GROES, and theoretical models of DNA conformational statistics.

Speaker's Bio:
Gregory S. Chirikjian received the B.S.E. degree in engineering mechanics, the M.S.E. degree in mechanical engineering, and the B.A. degree in mathematics, all from The Johns Hopkins University, Baltimore, MD, in 1988. He then received the Ph.D. degree from the California Institute of Technology, Pasadena, CA, in 1992. Since the summer of 1992, he has been with the Department of Mechanical Engineering, Johns Hopkins University, where he is now a Professor. Dr. Chirikjian is a 1993 NSF Young investigator, a 1994 Presidential Faculty Fellow, and a 1996 recipient of the ASME Pi Tau Sigma Gold Medal. His research interests include the kinematic analysis, motion planning, design, and implementation of biologically inspired robots. In particular, ``hyper-redundant,'' ``metamorphic,'' and ``binary'' manipulators, and most recently self-replicating robots. In recent years Dr. Chirikjian has also been applying methods from robotics to model conformational transitions in biological macromolecules.

Affiliation(s): Professor of Mechanical Engineering at Johns Hopkins University

Abstract:
The concept of self-replicating lunar factories was studied by NASA more than 2 decades ago, but until recently work in this area has been relatively dormant. Such factories would spawn lunar production of materials and energy on a massive scale and would dramatically impact man's ability to explore and colonize space.
Our concept of a self-replicating robotic factory consists of four subsystems: (1) multifunctional robots for digging and transportation of materials, and assembly of components during the replication process; (2) materials refining and casting facility; (3) solar energy conversion, storage and transmission; (4) Electromagnetic rail guns for long-distance transportation (e.g., for sending materials to low-Earth orbit (LEO), or transporting factories to distal points on the moon). Each of these subsystems is described in the context of current technologies, with an emphasis on (1).

The concept and history of self-replicating systems are reviewed. We build on previous concepts for self-replicating systems, present a simple prototype that demonstrates active mechanical self-assembly, and perform an analysis of how the proliferation of such systems on the lunar surface would occur.

Affiliation(s): Manager of Scientific Research in VISX, Incorporated, Santa Clara

Abstract:
Eye movements are an important aspect of human visual behavior. The temporal and space-variant nature of sampling a visual scene requires frequent attentional gaze shifts, saccades, to fixate onto different parts of an image. Experimental evidence suggests that fixations are often directed towards the most informative regions in the visual scene. We develop a model and its simulation that can select such regions based on prior knowledge of similar scenes. Having representations of scene categories as a probabilistic combination of hypothetical objects, i.e., prototypical regions with certain properties, it is possible to assess the likely contribution of each image region to the successive recognition process. Using conditional probabilities for each region given the scene category, the model can then predict its informative value and initiate a sequential spatial information-gathering algorithm analogous to an eye movement saccade to a new fixation. Selective expansion of information, plus the significant space-variant compression used, could allow for real-time scene analysis, for example, applications where a vehicle labels the scenes it comes across on its path even when the scenery changes very quickly. The algorithm establishes the most likely scene category for a given image.

Speaker's Bio:
Dr. Dimitri Chernyak is a manager of scientific research in VISX, Incorporated, the leading manufacturer of laser vision correction systems. He is actively involved in research on image feature registration and tracking and various optical wavefront technologies. Dr. Chernyak has received his PhD in Vision Science from UC Berkleley, specializing in research on top-down model-driven eye movements, visual attention, object recognition and image segmentation.

Title: "Towards Long-Term Autonomy in Embedded Systems Hybrid Modeling, FDI, Fault-Adaptive Control, and IVHM"

Affiliation(s): Department of Electrical Engineering and Computer Science Institute for Software Integrated Systems (ISIS) Vanderbilt University Nashville, TN

Abstract:
My research work in the last seven years has focused on modeling and analysis of complex systems, and its application to online fault detection and isolation, and fault-adaptive control. In the first part of my talk, I will briefly review the work we have performed in (i) modeling of complex, embedded systems, (ii) fault detection and isolation from transients in the behavior of dynamic systems, the TRANSCEND system, and (iii) fault-adaptive control using reconfiguration methods and online supervisory control. Our comprehensive diagnosis scheme combines monitoring, fault detection, fault isolation, prediction, and hypothesis refinement using an innovative qualitative framework that overcomes shortcomings of traditional quantitative schemes. The TRANSCEND scheme has been extended to the diagnosis of hybrid systems. Hybrid observers, fault detection schemes, fault isolation and identification, and reconfiguration have been integrated in the FACT methodology for fault-adaptive control. We have successfully applied the FACT scheme to fault-adaptive control of fuel transfer systems in fighter aircraft. Currently, we are developing schemes for fault-adaptive control of ALS systems.

Recently, I have become interested in issues that relate to autonomous and reliable long-term peration of complex embedded systems. This requires finding innovative solutions for a number of research problems: (i) modeling long term behavior of complex, distributed systems, (ii) online monitoring, FDI, and prediction analysis, and (iii) decision analysis for control, maintenance, and planning. In the second part of my talk, I will present a computational architecture for addressing this problem, and discuss methodologies that will need to be brought together to solve the various tasks defined by the architecture.

Affiliation(s): Associate Editor of the IEEE Trans IP and a Lecturer and Fellow of Trinity College, Dublin

Abstract:
With the increasing proliferation of multimedia digital content, it has been recognized that it is increasingly difficult to access the information it contains. There is a need for organizing multimedia records in the same way that text can be manipulated via Google for instance. Applications can be found in both consumer and scientific contexts. Users of music databases may find it easier to retrieve songs by humming rather than remembering the name, home viewers may wish for personalized content based summaries of Tennis (for instance) which shows all the famous people in the audience, Oceanography scientists pouring over footage from underwater cameras would like to access only the "interesting bits" without having to view the endless hours of open ocean in between. This topic of Video and Image retrieval has been an active area of research for some time. One of the main difficulties with work in this area is that without access to a user context there is no mechanism for determining how the feature space should be used to create a useful, content aware application.

The Signal Processing and Media Applications group (http://www.mee.tcd.ie/~sigmedia) at Trinity College, Dublin has been investigating retrieval with a view to exploiting context wherever possible. This talk introduces manly work in the domain of sports footage. This work exploits a new moment based feature that is able to extract court views (e.g. in Tennis and Snooker) without the need for 3D information. A combination of the HMM for parsing camera views, and a particle filter for ball tracking is also presented. Aspects of the use of audio and video for finding tennis services are also discussed. Finally, some emerging work in the area of parsing video from Psychological Surveillance experiments (for assessing the ability to execute particular motions) is presented.

Speaker's Bio:
Dr. Anil Kokaram is an Associate Ed. of the IEEE Trans IP and a lecturer and Fellow of Trinity College, Dublin. He was previously at the CUED working on Motion Picture Restoration and now leads the Signal processing and media applications group at Trinity College, Dublin http://www.mee.tcd.ie/~sigmedia. Research areas include multimedia content based retrieval for sports broadcasting and analysis, motion picture restoration, video over wireless and robust motion estimation.

Title: "PANEL: New Roles for Conferences, Journals, and Online Publications in Computing and Information Sciences: Making Research Results Timely, Authoritative, and Widely Accessible"

Abstract:
Scientific publications take many forms: informal "white papers", technical reports, workshop and conference articles, invited or refereed journal articles, book chapters, books, software/databases, and websites.

What are the various merits of each of these forms of publication? How do (or should) they factor into the evaluation of working scientists for hiring, retention, and promotion? How are various publications viewed by peer groups as shown in the awarding of grants and the conferring of awards? These are perennial issues in science.

Computer scientists often face particular challenges since many prestigious venues are in fact conferences (e.g. SIGGRAPH). While many CS conferences are stringently reviewed, conference publications are sometimes devalued by larger review bodies (e.g. tenure review, grant review panels) relative to journal articles. Unfortunately, the publication timeframe of some journals may not support the appearance of new results in a timely manner.

This panel will address these questions drawing on the publication practices of the panelists, and engage the audience in a discussion of the role of various forms of publications, both now and in the future.

Finally, we'll touch briefly on issues of visibility to upper corporate, government, and academic management, and outreach to the public.

Affiliation(s): Professor of ECE at University of Miami, Underwater Vision and Imaging Lab (UVIL)

Abstract:
Construction of accurate 2-D mosaicing and 3D terrain reconstruction systems by processing monocular and stereo flyover imagery is an important capability for a large number of terrestrial and underwater applications. A fundamental problem is to determine the motion and (or) trajectory of the camera platform in order to integrate images or depth maps from various views. Despite significant work on the development of accurate and efficient algorithms for motion estimation from image sequences, application to data sets recorded over large terrains result in significant drift errors. The first part of this talk covers some recent work on the application of differential flow-based methods for motion and 3-D depth recovery from video, and the application of a global alignment method to minimize drift error. The second part of the talk will cover some ongoing work on the real-time application of vision technologies for positioning and navigation of underwater remotely operated vehicles. This includes some recent work in applying stereovision for ship hull inspection.

Affiliation(s): Managing Director of FriarTuck Pte Ltd and Assistant Professor in the School of Computing, National University of Singapore

Abstract:
Constraint Programming provides an attractive framework for solving a wide range of event scheduling problems. Constraint programming systems provide abstractions (sometimes languages) or constraint-based problem modeling and software libraries (sometimes runtime systems) for solving constraint problems. The strength of the constraint programming paradigm stems from the fruitful interaction of three classes of algorithms; propagation algorithms, branching algorithms and exploration algorithms. This talk illustrates the problem solving capabilities of the constraint programming paradigm using the example of tournament scheduling. Domain specific global constraints employing graph matching enable significant performance improvements over competing technologies in this domain, including new results on established benchmarks in tournament scheduling.

The research results on global constraints for tournament scheduling were obtained in collaboration with Tobias Mueller (currently European Patent Office) and Sven Thiel (Max Planck Institute for Computer Science, Germany).

Speaker's Bio :
Dr. Henz studied Mathematics and Computer Science at the Saarland University, Saarbruecken, Germany. He obtained his MSc degree from the State University of New York at Stony Brook and his doctoral degree from the Saarland University. As research scientist at the German Research Center for Artificial Intelligence, DFKI, and the Department of Computer Science at the Saarland University, he contributed to the design and implementation of the concurrent constraint language Oz. Dr. Henz is a staff member (Assistant Professor since 1999 in the School of Computing at NUS, where he conducts research on combinatorial search problems, constraint programming system design and implementation, and teaches courses on programming languages and systems. His current research focuses on software engineering aspects of constraint programming systems. Martin is co-founder and Managing Director of FriarTuck Pte Ltd, a software company specializing in event scheduling. Dr. Henz has been serving as member of the programming committee of constraint programming and timetabling conferences, including the leading conference in constraint programming, CP2002, and the leading timetabling conference PATAT2000, PATAT2002, as well as MISTA2003. His papers on constraint programming applications appeared in Operations Research, European Journal of Operational Research and CP2001.

Title: "The Deductive Combination of Geospatial Resources on the Web to Answer Natural Language Queries"

Abstract:
CIP is being developed here at Ames for use at JPL during the next Mars rover mission. It will allow NASA researchers worldwide to access, display, and analyze rover commands uploaded to the rovers and scientific data downloaded from the rovers.

Title: "SOAREX Sub-Orbital Flight Experiments - A 'Wind-Tunnel in the Sky"

Abstract:
Researchers in the area of hypersonic flight experimentation and design face a peculiar dilemma. While there are now better analytical and design tools than have existed in the past, the discipline is conducting or completing fewer flight experiments. One of the approaches may be to find a manner in which the larger flight programs (X-vehicles) may be augmented by smaller, less expensive and focused flight experiments. The use of the NASA sounding rocket program is suggested as an attractive path. The advent of recently developed, light-weight MEMS (Micro-Electronic Machined Sensors) devices allows for experiment packages to be constructed at modest cost. Illustrating this methodology, two recent flights of the SOAREX (Sub-Orbital Aerodynamic Re-entry EXperiments) project have flown a total of 14 hypersonic experiments. These have included new planetary probe concepts as well as experiments involving high L/D 'waverider' designs. The discussion will describe the challenges of designing the rocket payload systems (the notion of extending the idea of a wind tunnel to a hypersonic flight test-bed) as well as the experimental methodology and results.

Speaker's Bio:
Marc Murbach is a member of the Advanced Projects Branch in the Space Projects Division. He is the originator and principal investigator of the SOAREX flight system and the associated experiments. His interests involve experimental hypersonic flight with application to future planetary missions as well as next generation reusable launch systems. His education includes Harvey Mudd College, San Jose State University, and Stanford University.

Abstract:
This lecture provides an introduction to Support Vector Machines, Kernel Principal Component analysis as examples for successful kernel-based learning algorithms. At first we give a short background about VC theory and kernel feature spaces and then proceed to kernel based learning in supervised and unsupervised scenarios including some practical and algorithmic considerations. If time permits, our recent work on kernel-based nonlinear source separation will be discussed.

We illustrate the usefulness of kernel algorithms by finally presenting applications such as DNA and biomedical data analysis (BCI).

Title: "Creating a Hypersonic Wind 'Tunnel-in-the-Sky' and Advanced High Speed Flight Research"

Abstract:
No doubt, the RoboCup challenge to beat the human soccer world champion by 2050 is well-known. The RoboCup initiative is aimed to foster AI research in the field of cooperative autonomous agents. A substantial part of Philips Center for Industrial Technology (Philips CFT) is devoted to mechatronical systems for fast and accurate positioning. The most appealing example of this field is probably the so called wafer stepper in which substrates move with a speed of up to 2 m/s maintaining accuracies down to 2 nm. A few years ago, Philips CFT decided to get involved in RoboCup, more precisely in the Middle Size League which is also known as the Engineering League. In the lecture, the mechatronic equipment design approach to a robot soccer challenge will be illustrated, focussing on the design and control of the motion engine, the kicking device and the machine vision system. Without question, our approach is much different from the one that most university teams have chosen, but provides some interesting viewpoints in relation to what one could call a "low-noise design". Systems with noisy sensors and actuators are extremely difficult to control towards a certain objective, even with more advanced intelligent control; reducing the noise then is the obvious choice.

Speaker's Bio:
Studied electrical engineering at Eindhoven University of Technology. Master's thesis in electromechanical actuator design. After graduation in April '92, started working at Philips Center for Industrial Technology, in a department called Mechatronic Systems. Worked on electromechanical actuator design, kinematic calibration of manipulators, wafer substrate stage design for both optical data drives, electron microscopes and semiconductor equipment. Latest involvement before moving to US was to lead a group on systems engineering to find ways of improving the effectiveness of our way-of-working. Currently located in Sunnyvale as manager of a small Mechatronic Systems group. Got involved in robocup on personal interests in December '99. Was project leader until move to US, now remotely involved in the project.

Abstract:
The fractal protein is a new concept intended to improve evolvability, scalability, exploitability and provide a rich medium for evolutionary computation and evolutionary design. Here the idea of fractal proteins and fractal proteins with concentration levels are introduced, and a series of experiments showing how evolution can design and exploit them within gene regulatory networks is described. These show how fractal GRNs have a natural tendency towards damage tolerance, and can be successfully employed for tasks such as robot control.

Patterns are crucial in development. Biological development is not simply a process to create a phenotype. It is an on-going process that shapes, adapts and repairs an organism, beginning at conception and ending at the death of the organism. Consequently, an organism is not an end-result of a developmental program, it is an ever-changing solution that is maintained by development. From a computational viewpoint, the processes of development appear to be constant, never-ending patterns, generated by gene-gene interactions in gene regulatory networks and at higher levels in cell-cell interactions, and so on. Phenotypes are the dynamic high-level results of those non-stop lower-level interactions.

Speaker's Bio:
Dr. Peter J. Bentley is an Honorary Research Fellow at the Department of Computer Science, University College London (UCL), and Visiting Fellow of University of Canterbury, known for his research covering all aspects of Evolutionary Computation and Digital Biology. His research investigates evolutionary algorithms, ecological modelling, artificial immune systems, computational embryology and swarming systems, applied to diverse applications including design, control, fraud detection, and security.

Abstract:
Planning natural language sentences requires generation systems to bridge underlying domain-specific representations and general linguistic representations. This is a complex task, in which the generator must organize related domain facts to fit linguistic structures, formulate linguistic expressions to identify domain entities, and select particular words and constructions to realize domain concepts. The task requires correspondingly complex resources, which must offer a fine-grained and comprehensive model of linguistic phenomena at syntactic, semantic and pragmatic levels.

This talk explores the consequences for system-building of a uniform approach to sentence planning based explicitly on representing and reasoning about the intended descriptive connections between the system's utterances and its environment. Adopting this approach makes generators more flexible. By using representations of intended meaning, a generator can augment the syntax, semantics and pragmatics of an incomplete sentence simultaneously, and can assess its progress on the various problems of sentence planning incrementally. Adopting this approach also makes generators easier to build. Intention representations are declarative combinations of linguistic resources. The close link allows us as designers to formulate grammatical and conceptual resources to achieve desired generator output, and even to create resources automatically that will allow a generator to reproduce marked-up sample utterances.

Speaker(s): Bruce Lewis, Experimental Developer and Chair of the SAE subcommittee AS-2C Architecture Description Language

Abstract:
The Society of Automotive Engineers, Aerospace Avionic Systems Division, is developing a standard architecture description language suitable for the avionics, space and other real-time domains useful for architecture analysis and model driven automated system integration. The AADL is a computer language used to describe the software and hardware components, the interfaces between the components and the dynamic behavior of the components. The language can describe important non-functional aspects such as timing requirements, fault and error behaviors, time and space partitioning, and safety and certification properties. It also supports code generation of the system executive and glue code to automate the integration of software and hardware components to the specified requirements. The AADL is based on capabilities demonstrated in the MetaH language and toolset. MetaH was developed by Honeywell under DARPA funding and US Army support and experimentation.

The language integrates the concerns of the system engineer and the software engineer. It supports system development from requirements trade-off analysis through component development and final integration and then across the life cycle for rapid system evolution. Tools can be used to predict the emergent behavior of the composed components such as schedulability, reliability, safety (as demonstrated with MetaH). The AADL will also include a standardized expression of the language as a UML profile allowing precise modeling and integration using UML tools. This language / technology / tools provide a significant means to accelerate the development of and to rapidly evolve real time safety critical systems.

The presentation will present an overview of the concepts and development approach that would be used with the AADL as demonstrated with the MetaH toolset. It will also provide an overview of standardization progress, the language, and current analysis approaches. Bruce will be available for individual appointments after the seminar. Please contact Julian Gomez 4-0770 email:jgomez@mail.arc.nasa.gov to get on his schedule.

Title: "Optimal Partitions for Density Estimation, Cluster Analysis, Signal Detection, and other Data Analysis Problems"

Speaker(s): Jeffrey D. Scargle, Space Science Division / Bradley W. Jackson, Department of Mathematics

Abstract:
Astronomers are making enormous breakthroughs in understanding the origin, evolution and nature of the Universe, largely due to data-rich observational programs from space, airborne and ground-based platforms. The corresponding information technology challenges include the development of statistical models capable of automated analysis of very large and complex data sets.

Toward this goal, we describe a novel method for detecting and characterizing signals in noisy time series and other sequential data. The fitness function we adopt for this nonparametric model is the marginal posterior, for which we give simple exact formulas for several data modes. We have solved the resulting combinatorial optimization problem with a dynamic programming algorithm that yields the exact, global, optimal partition of the data interval in time O(N**2), where N is the number of data points. We will also show preliminary results for extending the same segmentation ideas to 2D and higher. As an example, we show analysis of data from the Sloan Digital Sky Survey -- graphically depicting clusters and other groupings of galaxies distributed throughout the Universe.

Abstract:
HDCP is a NASA-funded five-year research program whose goals are to: 1) help create a rigorous foundation for software research; and 2) develop new techniques and technologies for improving the dependability of NASA systems. In this talk I will overview the program and explain how we are approaching those goals.

Speaker's Bio:
Michael Evangelist is a professor of engineering practice within the School of Computer Science at Carnegie Mellon University and director of research for the West Coast campus. He has spent most of his career in industrial research at Bell Labs, MCC, and Andersen Consulting – and recently was director of the Computer-Communications Research Division at the National Science Foundation. His research interests include dependable software engineering, distributed computing, and theoretical computer science.

Title: "Query Processing, Resource Management, and Approximation in a Data Stream Management System"

Abstract:
This talk will describe our ongoing work developing the Stanford Stream Data Manager (STREAM), a system for executing continuous queries over multiple continuous data streams. The STREAM system supports a declarative query language, and it copes with high data rates and query workloads by providing approximate answers when resources are limited. In addition to motivation, overview, and research directions, we will highlight several specific contributions to date: syntax and formal semantics for continuous queries over streams and relations; structure of query plans in our system; scheduling, resource allocation, and approximation techniques to cope with limited resources; and exploiting constraints on data streams to reduce memory overhead.

Speaker's Bio:
Jennifer Widom is an Associate Professor in the Computer Science and Electrical Engineering Departments at Stanford University. She received her Bachelors degree from the Indiana University School of Music in 1982 and her Computer Science Ph.D. from Cornell University in 1987. She was a Research Staff Member at the IBM Almaden Research Center before joining the Stanford faculty in 1993. Her research interests include query processing on data streams, data caching and replication, semistructured data and XML, and data warehousing.

Abstract:
MIT has established the Engineering Systems Division to develop the intellectual foundation and curriculum for addressing large scale engineering systems with major societal impact. This talk will consider aircraft as an example of an Engineering System. A framework of three orthogonal dimensions - Technical, Social and Lifecycle - each with substructure, characterize aircraft systems. Interrelationships within and between the substructure of these dimensions, and among the dimensions, are critical as will be illustrated by numerous examples. A value creation framework is introduced to for development of aircraft to create of lifecycle value and is illustrated with case studies of existing aircraft (F-16, B-777, F/A-18, JAS-39).

Speaker's Bio:
Earll Murman is Ford Professor of Engineering in the Department of Aeronautics and Astronautics and the Engineering Systems Division at MIT. He served as Head of MIT's Department of Aeronautics and Astronautics from 1990-1996, as Director of Project Athena from 1988-1991, and as Co-Director of the Lean Aerospace Initiative from 1995-2002. In addition to his 22 years in academia, his career includes 4 years at the Boeing Company, 6 years with Flow Research Co., and 3 years at NASA's Ames Research Center. Dr. Murman's professional interests span system engineering, product development, lean manufacturing, aerodynamics, computational fluid dynamics, and engineering education. He has published 90 journal articles and technical papers in these fields. Dr. Murman graduated summa cum laude in Aeronautical Engineering from Princeton University in 1963 and received his PhD in Aerospace Engineering from Princeton in 1967. He is a member of the U.S. National Academy of Engineering and a Fellow of the American Institute of Aeronautics and Astronautics and the Royal Aeronautical Society.

Speaker(s): David Draper, Athanasios Kottas, Herbie Lee, Raquel Prado, Bruno Sanso

Abstract:
In this joint talk we will each provide 10-minute summaries of some recent developments in Bayesian statistical modeling and analysis on which we concentrate in our research. Methodology topics to be covered -- all from the Bayesian point of view -- will include hierarchical modeling, inverse problems, machine learning, Markov chain Monte Carlo methods, model uncertainty, nonparametrics, nonstationary time series, signal processing, spatio-temporal modeling, stochastic optimization and survival analysis, and application areas will include biomedical monitoring, environmental studies, quality assessment in health and education, rainfall prediction, and risk assessment.

Speaker's Bio:
Professor David Draper Chair, Department of Applied Mathematics and Statistics Baskin School of Engineering
University of California
1156 High Street, Santa Cruz CA 95064 USA
http://www.ams.ucsc.edu/~draper/
draper@ams.ucsc.edu
phone US (831) 459 1295, nonUS +1 831 459 1295
fax US (831) 459 4829, nonUS +1 831 459 4829
departmental web pages: http://www.ams.ucsc.edu

Title: "What I learned at the Sloan Program (or how I spent my year at Stanford)"

Abstract:
Every year NASA has the option to send one Civil Servant to the Stanford Sloan Program. In 2002, David Korsmeyer was selected to attend. The Sloan Program is full-time 1 year academic program for mid-career managers leading to a Master of Science in Management degree (to some an oxymoron) from the Stanford Graduate School of Business. This includes core courses for the Sloan Program as well as 2nd year MBA and PhD electives. Some of the lessons and insights from the Program will be discussed as will pertinent impacts upon NASA and work within Code IC.

Speaker's Bio:
Dr. David Korsmeyer is the Code IC Assistant Division Chief for Programs, and the Collaborative Assistant Systems (CAS) Technical Area Lead. He has recently returned from the Stanford Sloan program. Dr. Korsmeyer obtained his PhD in Aerospace Engineering from the University of Texas in 1991 and has been at NASA ARC since.

Abstract:
Over the years, the research and development of autonomy approaches for their use in terrestrial and space robotic applications has been one of the most exciting and pressuring fields of activity in space technology. Obviously, the use of sensory subsystems and sensory data processing plays a key role in conceptualizing this envisioned, increased level of capability, notably vision and navigation processing in real-time. On the other hand, these sensory subsystems need to be integrated to advanced control, navigation, and communications subsystems, if the space robots are to operate in an autonomous fashion. The speaker has designed and built space robots and led associated space robotic missions (real-time vision and control) for the orbital infrastructure and deep space contexts.

More revolutionary approaches with high potential to solve long-standing problems in vision, control, and navigation of autonomous robots have been recently and continue being investigated. Their biological computing inspiration is notable. In this invited presentation, the status quo of this new thrust of approaches is reviewed, making emphasis in the detail presentation of the most advanced biological-inspired computing architectures, the comparison to current approaches to solve the same or similar tasks, and the description of some system prototypes including flying robots based on these revolutionary computing approaches.

Speaker's Bio:
Dr. V. David Sánchez A. is the President and CEO of California Aerospace Enterprises, Inc., a company specializing in the research and technology development of aerospace technologies and systems. Dr. Sánchez has held over twenty years leadership positions in the U.S. and Europe in the areas of aerospace and intelligent systems in national labs, industry, and academia including a Chief Technology Officer position at Thuris Corp. in Newport Beach, CA, a company specializing in the application of advanced brain research results and intelligent computational methods, senior Computer Science, Engineering, and Technology faculty positions at the University of Miami, Coral Gables, Florida and Devry University, California , and a senior research scientist position for almost a decade at the German Aerospace Research Establishment (DLR), Oberpfaffenhofen, Germany. He has held numerous keynote speaks and invited presentations all over the world and at most of the major NASA and ESA labs, is author of many publications in the areas of robotics, computer vision, neurocomputing, pattern recognition, control, space and brain research. He has been involved in most sectors of the aerospace endeavor including orbital infrastructure (satellite, spaceshuttle, spacelab) and deep space (planets, moons, asteroids) missions, research, and technology development including the use of computational intelligent, biologically-inspired approaches and systems in these domains. Dr. Sánchez founded in 1989 and since then has served as the Editor-in-Chief of Neurocomputing, an Elsevier Science journal, vol. 49-55 to be published in 2003.

Abstract:
Program transformation, i.e., automatically modifying programs according to some set of rules, is a core technology in automated software engineering and has applications ranging from compilation, application generation and program synthesis, to optimization, software renovation, and reverse engineering. In order for program transformation to become a standard tool in every software engineers toolbox, support for concise and high-level, executable specification of program transformations is necessary.

Stratego is a language for program transformation aimed at providing support building transformation systems. The language is based on the paradigm of term rewriting with programmable rewriting strategies. Features of the language include conditional rewrite rules for specifying basic transformation steps, strategies for controlling the application of rules, basic combinators for building generic traversals over syntax trees, dynamic rewrite rules for dealing with context-sensitive transformations, concrete object syntax for specifying transformations in the syntax of the programming language at hand. The Stratego compiler translates specifications to C code. The Stratego Standard Library comes with a large base of generic and domain-specific transformation strategies.

The XT bundle of program transformation tools is a tool suite built around Stratego and the syntax definition formalism SDF. XT provides support for parsing, pretty-printing, syntax tree manipulation, and grammar engineering, and includes a grammar base with syntax definitions for many languages. Stratego and XT are open source under the GPL licence.

In this talk I will give an overview of the main features of Stratego illustrated with examples.

Speaker's Bio:
Dr. Visser holds a Doctorate in Computer Science from the University of Amsterdam, completed in September 1997. His dissertation was Syntax Definition for Language Prototyping. He has participated in research in: Models and Techniques for Variability Management, SERC,. A Transformation Factory for Digital Signal Processing Software., and the generation of program transformation systems.

He has developed software including: Stratego, a language for program transformation based on the paradigm of rewriting strategies, SDF2, a formalism for the definition of the syntax of computer languages, Grammar Tools,a collection of tools for grammar analysis, grammar (re)construction, and tree manipulation, transformation tools, Haskell Transformation, CodeBoost?, a source-to-source transformation system for C++ programs, and CobolX?, a system for source-to-source transformation of Cobol programs with the purpose of renovation of legacy code.

Eelco Visser
Institute of Information and Computing Sciences
Universiteit Utrecht
P.O. Box 80089
3508 TB Utrecht
The Netherlands
visser@acm.org
http://www.cs.uu.nl/~visser/
Phone: +31 30 253 4592
Fax: +31 30 251 3791

Title: "Formal Verification of Diagnosability via Symbolic Model Checking"

Abstract:
This paper addresses the formal verification of diagnosis systems. Given a physical system and a diagnosis system that observes it, we tackle the problem of verifying that the diagnosis system will be able to recognize the situations that it is required to recognize. In our approach, the physical system is formally modeled as a Kripke structure, while diagnosability is checked by looking for pairs of scenarios that (1) are indistinguishable based on the observable part of the system but (2) lead to situations that are required to be distinguished. We focus on the practical applicability of the method: diagnosability is recast in terms of a model checking problem, and allows for the direct use of state-of-the-art symbolic model checking techniques.

Speaker's Bio :
Charles Pecheur RIACS / ASE Group
NASA Ames Research Center, M/S 269-2 Bldg N-269, Rm 260-14
Moffett Field, CA 94035, U.S.A. http://ase.arc.nasa.gov/pecheur
Phone : +1-650-604-3588 pecheur@ptolemy.arc.nasa.gov
Fax : +1-650-604-3594