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

RIACS Seminar #50

Date: December 4, 2000

Title: "Model Checking of Autonomy Models for a Mars In-Situ Propellant Production Facility"

Speaker(s): Peter Engrand

Affiliation(s): NASA Kennedy Space Center

Abstract:
This is a preview of a presentation to be given at the RIACS Workshop on V&V of Autonomous and Adaptive Systems on Wednesday. The talk relates the ongoing efforts at KSC in using the SMV model checker to support the development of an autonomous controller based on the Livingstone system for the Mars-bound ISPP facility. It gives concrete examples of the kind of errors that have been addressed using V&V, and illustrates the difficulties met by a developer not acquainted with formal methods when trying to use those tools on real problems.

The ISPP (In-Situ Propellant Production facility) is designed to produce spacecraft propellant out of the CO2 in Mars atmosphere. Livingstone is a model-based diagnosis system developed by the Autonomy group here at Ames. SMV is a symbolic model checker from Carnegie Mellon University. A translator from Livingstone models into SMV models has been developed by the ASE group here at Ames, in partnership with CMU.

RIACS Seminar #49

Date: November 30, 2000

Title: "Inside a Minute and a Half: Participant Observation in Work Systems Design"

Speaker(s): Zara Mirmalek

Affiliation(s): RIACS, NASA Ames Research Center

Abstract:
1.5 mintues is an observed length of time for the transaction that takes place between an employee and a customer at United Airlines. An examination of the technology and human factors that are brought together for these moments to occur are looked at using the multiple perspectives available in the method of participant-observation. In an ethnogrpahy of customer service representatives and customers at United Airlines data was collected for considerations of changes to design of the work system in the areas of training and application.

RIACS Seminar #48

Date: November 9, 2000

Title: "Learning Low Level Actions for Robot Navigation"

Speaker(s): Dr. Larry Pyeatt

Affiliation(s): Texas Tech University

Abstract:
This seminar presents an architecture for mobile robot navigation in which low level actions are learned on-line as the robot performs its tasks. The actions are adaptive to failures in sensors and effectors, allowing the robot to perform its assigned tasks despite hardware failure. Reactivity, deliberation, and learning are an integral part of the architecture. The architecture uses a partially observable Markov decision process (POMDP) approach for path planning, and reinforcement learning (RL) for low level actions. The initial implementation of this architecture has been validated using simulation. The experiments conducted using simulation indicate that the combination of POMDP planning and reinforcement learning provides a very reactive system that can also achieve long term goals, adapt to failures, and learn new low level actions. Work is currently in progress to port the architecture to run on a mobile robot.

RIACS Seminar #47

Date: November 9, 2000

Title: "Automatic Parallel Control Structures in SequenceL"

Speaker(s): Dr. Daniel E. Cooke

Affiliation(s): Texas Tech University and NASA Ames Research Center

Abstract:
SequenceL is a language that provides declarative constructs for nonscalar processing. Rather than specifying program control structures that, in turn, imply a data product, the problem solver specifies a data product and the control structures to produce or process the data product are implied. Although SequenceL has been previously introduced in two papers, recent improvements to the language have indicated that parallel control structures are also implied by the SequenceL problem solutions. This presentation focuses on these recent advances.
The talk will be an overview of the execution of SequenceL functions for Matrix Multiply, Gaussian Elimination, and Quicksort - and compare the SequenceL solutions with solutions provided in multithreaded JAVA and C/MPI codes. We are confident that the current research results will allow one to find the inherent (or natural) parallelisms that exist in a given problem solution. The examples reviewed in this talk indicate SequenceL's adeptness in handling a good range of parallel problems. The matrix multiply is an example where there are no dependencies among the computations taking place in the parallel paths. The Gaussian Elimination code is one where there are dependencies among the parallel computations. In terms of scheduling, both the matrix multiply and the Gaussian Codes are examples of problems for which static a-priori schedules can be generated. The paths of execution can be determined based upon the dimensions of the matrix, in the matrix multiply, and the number of equations, in the Gaussian code. The final example, the Quicksort, provides insight into problems requiring dynamic scheduling - the number and schedule for the parallel paths occurs on the fly, regardless of the method used to choose the pivot. The talk will then move towards the facility SequenceL provides in new exploration paradigms. In these new paradigms, processes seek out data rather than having data presented to processes in a traditional input mechanism. Furthermore, processes share state information so that they can pass around partial computations.

RIACS Seminar #46

Date: October 30, 2000

Title: "Bounded Rationality, $-Calculus and Power Grid"

Speaker(s): Eugene Eberbach

Affiliation(s): Professor,Principal Scientist, Jodrey School of Computer Science Acadia University, Canada

Abstract:
Recently, there has been shift from consideration of optimal decisions in games to a consideration of optimal decision-making programs for dynamic, inaccessible, complex environments such as the real world. Perfect rationality is impossible in these environments, because of prohibiting deliberation complexity.

Anytime algorithms attempt to trade off result quality for the time or memory needed to generate results. Bounded rational agents are ones that always take the actions that are expected to optimize their performance measure, given the percept sequence they have seen so far and limited resources they have. Process algebras, with basic programming operators, has been used to study the behaviors of interactive multi-agent systems and leading to more expressive models than Turing Machines, e.g., Interaction Machines.

By extending process algebra operators with von Neumann/Morgenstern's costs/utilities, anytime algorithms can be viewed as a basis for a general theory of computation. As the result we shift a computational paradigm from the design of agents achieving one-time goals, to the agents who persistently attempt to optimize their happiness. We call this approach $-calculus (pronounced "cost-calculus"), which is a higher-order polyadic process algebra with a utility (cost) allowing to capture bounded optimization and metareasoning in distributed interactive AI systems.

$-calculus extends performance measures beyond time to include answer quality and uncertainty, using k Omega-optimization to deal with spatial and temporal constraints in a flexible way. This is a very general model, just as neural networks or genetic algorithms, leading to a new programming paradigm (cost languages) and a new class of computer architectures (cost-driven computers). The NSERC supported project on $-calculus aims at investigation, design and implementation of a wide class of adaptive real-time distributed complex systems exhibiting meta-computation and optimization. It has also been applied to the Office of Naval Research SAMON robotics testbed to derive GBML (Generic Behavior Message-passsing Language) for behavior planning, control and communication of heterogeneous Autonomous Underwater Vehicles (AUVs). Some preliminary ideas have also been utilized in the 5th Generation ESPRIT SPAN project on integration of object-oriented, logic, procedural and functional styles of programming in parallel architectures.

It appears that $-calculus can be useful for the NASA Information Power Grid (IPG) Project. The IPG testbed provides access to a widely distributed network of high performance computers. $-calculus resource-bounded optimization allows for flexible allocation of resources and scalability needed to tackle hard computation problems, thus $-calculus could provide a unifying metasystem framework for the Information Power Grid.

Speaker's Bio:
Dr. Eberbach is a Professor at School of Computer Science, Acadia University and an Adjunct Professor at Faculty of Graduate Studies, Dalhousie University, Canada. Previously he was Senior Scientist at Applied Research Lab, The Pennsylvania State University, Visiting Professor at The University of Memphis, USA, Research Scientist at University College London, U.K., Assistant Professor in Poland, and he also has industrial experience. Professor Eberbach's current work is in the areas of process algebras, resource bounded optimization, autonomous agents and mobile robotics. General topics of interest are new computing paradigms, languages and architectures, distributed computing, concurrency and interaction, evolutionary computing and neural nets. More information about projects, publications, courses taught can be found at http://cs.acadiau.ca/~eberbach.

RIACS Seminar #45

Date: October 27, 2000

Title: "Smart Sensors, Collaborative Sensemaking"

Speaker(s): Dr. Feng Zhao

Affiliation(s): Principal Scientist, Xerox PARC

Abstract:
Imagine a world in which we live where smart roads would be able to tell us when they need repair and which is the best direction to get to the Giants game, smart factories would stock up just enough inventory, ... The rapid advances in micro-electro-mechanical systems (MEMS) and lower-power wireless networking have enabled a new generation of tiny, cheap, networked sensors that can be "sprayed" on roads, across machines, and on walls. However, these massively distributed sensor networks must overcome a set of technological hurdles before they become widely deployable. Keeping up with the constant onslaught of sensory data from say 100,000 sensors is akin to drinking from a fire hose.

The Xerox PARC Smart Matter Diagnostics and Collaborative Sensing Project studies the fundamental problems of distilling high-level, human-interpretable knowledge from distributed heterogeneous sensor signals in a rapid and scalable manner. We are developing powerful algorithms and software systems to enable a wide range of applications, from sensor-rich health monitoring of electro-mechanical equipment to human-aware environments that leverage sensors to support synergistic interactions with the physical world.

Speaker's Bio:
Feng Zhao is a Principal Scientist in the Systems and Practices Laboratory at Xerox PARC. Dr. Zhao leads the Smart Matter Diagnostics Project that investigates how sensors and networking technology can change the way we build and interact with physical devices and environments. His research interest includes distributed sensor data analysis, diagnostics, qualitative reasoning, and control of dynamical systems.

Dr. Zhao received his PhD in Electrical Engineering and Computer Science from MIT in 1992, where he developed one of the first algorithms for fast N-body computation and phase-space nonlinear control synthesis. From 1992 to 1999, he was Assistant and Associate Professor of Computer and Information Science at Ohio State University. His INSIGHT Group developed the SAL software tool for rapid prototyping of spatio-temporal data analysis applications; the tool is currently used by a number of other research groups. Currently, he is also Consulting Associate Professor of Computer Science at Stanford.

Dr. Zhao was National Science Foundation and Office of Naval Research Young Investigator, and an Alfred P. Sloan Research Fellow in Computer Science. He has authored or co-authored about 50 peer-reviewed technical papers in the areas of smart matter, artificial intelligence, nonlinear control, and programmming tools.

RIACS Seminar #44-A

Date: October 12, 2000

Title: "Influence of Variations on Systems Performance and Safety"

Speaker(s): Irem Tumer

Affiliation(s): NASA Ames Research Center, Intelligent Health and Safety Group

Abstract:
High-risk aerospace components have to meet very stringent quality, performance, and safety requirements. Any source of variation is of concern, as it may result in scrap or rework (translating into production delays), poor performance (translating into customer dissatisfaction), and potentially unsafe flying conditions (translating into catastrophic failures). As part of the Intelligent Health and Safety group, we have been designing controlled experiments to understand various sources of variations in helicopter transmissions, collecting vibration data, and analyzing the data for indicators of the variations. We are looking for normal and abnormal sources of variation that affect performance and indicators of these variations to provide warning about potential failures during flight. The experiments include:

1. Flight tests using an AH-1 and an OH-58 helicopter, to determine the variations introduced due to regular maneuvering and the covariance with environmental conditions, engine torque, etc.;
2. OH-58 tranmission test-rig tests to determine the effect of variations due to different levels of torque, mast bending, and mast lifting forces, as well as pinion reinstallation effects;
3. Machinery Fault Simulator tests to test the effect of prefabricated defects and inherent design and manufacturing variations on gears, bearings, etc.

In this talk, I will present an overview of our group's research goals, discuss the experiments and go over some of the results from the data analyses conducted so far. I will then discuss the current work and future directions in developing formalized methods for design and manufacturing engineers, using the variation information from empirical and analytical studies.

RIACS Seminar #44

Date: September 28, 2000

Title: "Motion Strategies for Autonomous Observers"

Speaker(s): Hector H. Gonzalez-Banos

Affiliation(s): Stanford University, Computer Science Robotics Lab

Abstract:
This talk presents techniques to generate motion strategies for a new class of autonomous agent called the Autonomous Observer. An autonomous observer (AO) is a physical agent performing high-level vision-oriented operations, such as tracking a moving target or building maps of an environment. What distinguishes an AO from other autonomous agents is the use of its sensors as end-effectors. In traditional robotics, sensing is a means to an end --- e.g., sonars are used for collision avoidance in robot navigation, cameras are used to recognize individual parts in assembly tasks, and proximity sensors enhance grasping operations. For an AO, on the other hand, information gathering is the goal. Building models of objects and/or environments, detecting faults in large structures, tracking moving targets, or performing surveillance operations, are all examples of tasks for AO systems.

One characteristic feature of all AO-based systems is the need to satisfy geometric visibility constraints while planning and executing motions. Although similar problems have been studied in other contexts, the material presented in this talk focuses on the fundamental motion planning problem rather than on the vision or sensing aspects: Which locations must be visited by a robot to efficiently map a building? How must a robot proceed in order to explore an enviroment? What motions will keep a target in view? How can we reduce the number of sensing operations? To answer these questions we propose randomized art-gallery and next-best-view algorithms. These algorithms have been integrated in robot systems and experimental results will be described.

hhg@robotics.stanford.edu

RIACS Seminar #43

Date: September 14, 2000

Title: "Disfluency Studies at Telia Research"

Speaker(s): Robert Eklund

Affiliation(s): Telia Research & Linköping University

Abstract:
Automatic speech recognition (ASR) has reached a level of reliability that allows use in commercial applications. However, not all problems with regard to ASR have been solved, and one particular field that remains to be solved is the handling of 'disfluencies', typical of spontaneous speech, i.e., filler words like "eh", "uh", unfilled pauses, repeated words, truncated words, mispronunciations, prolonged segments and other hesitation phenomena.

This talk gives an introduction to ongoing disfluency studies at Telia Research, Sweden's biggest telecom operator. The talk will cover methodology and data collection as well as different speech corpora. Results from different corpora are presented, and comparisons are made between human--machine interaction using different modalities e.g. telephone conversations vs. graphical interfaces.

Some results from cross-linguistic studies (Swedish vs. American English; Swedish vs. Tok Pisin) are will also be presented.

RIACS Seminar #42

Date: September 14, 2000

Title: "Learning Probabilistic Relational Models"

Speaker(s): Lise Getoor

Affiliation(s): Stanford University

Abstract:
A large portion of real-world data is stored in commercial relational database systems. In contrast, most statistical learning methods work only with ``flat'' data representations. Thus, to apply these methods, we are forced to convert our data into a flat form, thereby losing much of the relational structure present in our database. I'll describe recent work on probabilistic relational models (PRMs), and describe how to learn these models directly from structured data. PRMs allow the properties of an object to depend probabilistically both on other properties of that object and on properties of related objects. PRMs are more expressive than standard graphical models, such as Bayesian networks, and I'll show how to extend well-known statistical methods for learning Bayesian networks to learn these models. I'll describe a range of learning algorithms, beginning with methods for learning models with uncertainty over attribute values, next incorporating class hierarchies, and finally learning algorithms that handle structural uncertainty. As we go along, I'll present experimental results on both real and synthetic data. Joint work with Nir Friedman, Daphne Koller, Avi Pfeffer and Benjamin Taskar

RIACS Seminar #41

Date: August 31, 2000

Title: "Holographic Optical Data Storage with Bacteriorhodopsin"

Speaker(s): Dogan A. Timucin

Affiliation(s): RIACS, NASA Ames Research Center

Abstract:
The holographic storage paradigm has been pursued for nearly four decades as an alternative to traditional memory technologies. This quest has followed a cyclical path between vitality and dormancy over time - a typical symptom of the disease of fashion in science: interest was fueled by the promise of unprecedented storage capacities and access times on one hand, while real progress was hindered by the lack of many enabling technologies on the other. In this (tutorial) talk, I will start by reviewing the principles behind holography and optical information processing. I will then describe holographic optical data storage architectures and materials, and discuss the current status of research and development efforts in this field. Finally, I will highlight a promising bio-optical material, bacteriorhopsin, with which we have worked recently toward building a holographic storage test-bed for NASA applications.

RIACS Seminar #40

Date: August 17, 2000

Title: "SSRP Presentations, Group 2"

Speaker(s) & Title(s):

Dan Bernstein of University of Massachusetts: "Rover Control as a Markov Decision Process"
Jeff Thompson of University of Minnesota: "Property Specification for Java Pathfinder"
Jason Baldridge of University of Edinburgh: "Integrating Natural Language Processing and Eye Tracking in Interfaces for Semi-autonomous Agents"
Vandi Verma of Carnegie Mellon University: "Non-parametric State and Fault Identification for Rovers"
Forrester Cole of Harvard University: "Implementation Issues with Haptics in an Immersive Virtual Environment"
Glen Nuckolls of U.C. Davis: "Incremental Reasoning in (non-binary) CSPs"

Abstract:
Students will discuss the specific problem or task(s) they are working on, the goals to be accomplished with this work, the approach taken to solve the problem or perform the tasks, the current status of the work and possible interesting future extensions to this work.

RIACS Seminar #39

Date: August 14, 2000

Title: "Spoken Dialogue Management using Dialogue Moves and Information States"

Speaker(s): Jim Hieronymus

Affiliation(s): Gothenburg University

Abstract:
Spoken dialogue systems which closely model human-human dialogues will be easy to use and more readily accepted by the general public. The EU Siridus project systems under development by Gothenburg University, SRI Cambridge, Telefonica and University of Seville attempt to model how humans behave in task oriented dialogues. Given a task domain, it is possible to define a set of dialogue moves which cover all of the utterances from a database of human-human dialogues on the same task. Information states provide a framework for implementing plans and flexible dialogue analysis to determine the present speaker move and the next move by the system. Example dialogues from the travel information domain will be shown, dialogue moves and information state structures used to manage a very natural, mixed iniative dialogue will be discussed. A new asynchronous implementation of the dialogue manager has just been implemented which allows processing of acknowledgement moves during system speech will be discussed and perhaps a demo will be shown of this system working.

RIACS Seminar #38

Date: August 11, 2000

Title: "Soft Constraints"

Speaker(s): Francesca Rossi

Affiliation(s): University of Padova, Italy

Abstract:
Soft constraints add to the classical notion of constraint the possibility of dealing with important features like fuzziness, uncertainty, optimization, probability, and partial satisfaction. This talk will describe the current state-of-the-art in the area of soft constraints, by reviewing the existing frameworks and pointing out the relations among them. Then, it will focus on one of the most general frameworks for soft constraints, which is based on a semiring structure, and, for such a framework, it will present its properties and constraint propagation algorithms. Notions of abstraction and learning for soft constraints will also be given. Finally, it will describe and show the usefulness of a logic-based programming language, called clp(fd,S), where soft constraints can be naturally used and are efficiently implemented.

frossi@math.unipd.it

RIACS Seminar #37

Date: August 10, 2000

Title: "SSRP Presentations, Group 1"

Speaker(s) & Title(s):

Kristin Branson of Harvard University: "Converting STRIPS Planning Domain Representations to Equivalent Remote Agent Representations"
Brian Murphy of Stanford University: "Towards Solving NP-Complete Problems with Quantum Computation"
Adrian Agogino of University of Edinburgh: "Collective Learning of Action Sequences in Groups of Semi-Autonomous Agents"
Ralph Benzinger of Cornell University: "Automated Analysis of Convergence Properties of Synthesized Programs"
Sergey Kirshner of U.C. Irvine: "Using Statistic Matching for Cross-validation Error to Improve Generalization"

Abstract:
Students will discuss the specific problem or task(s) they are working on, the goals to be accomplished with this work, the approach taken to solve the problem or perform the tasks, the current status of the work and possible interesting future extensions to this work.

RIACS Seminar #36

Date: July 26, 2000

Title: "Computational Models of Grounding in Collaborative Systems"

Speaker(s): Dr. David Traum

Affiliation(s): University of Maryland

Abstract:
Common ground (or mutual belief) between conversational participants is assumed to be crucial for many collaborative tasks. However, the process by which this common ground is augmented (called Grounding) has often been either oversimplified or studied in an off-line manner. I will present two computational approaches to grounding, using ideas from speech act theory, and discuss their advantages and short-comings for helping enable a computer system to analyze and engage in cooperative spoken and multi-modal communication. The first, from Traum 1994, views grounding related action as performing one or more of seven types of "grounding acts", and uses finite state automata to track the state of grounding and serve as context for planning grounding related utterances. The second, presented in Traum and Dillenbourg 1996, 1998, generalizes the notion of grounding act function, and uses a utility theory model to help select an appropriate next action.

References:
David R. Traum "A Computational Theory of Grounding in Natural Language Conversation, TR 545 and Ph.D. Thesis, Computer Science Dept., U. Rochester, December 1994

David Traum and Pierre Dillenbourg , Miscommunication in Multi-modal Collaboration, In working notes of the AAAI Workshop on Detecting, Repairing, And Preventing Human--Machine Miscommunication, pages 37-46, August 1996.

David R. Traum and Pierre Dillenbourg, Towards a Normative Model of Grounding in Collaboration, in working notes, ESSLLI-98 workshop on Mutual Knowledge, Common Ground and Public Information, August 1998.

RIACS Seminar #35

Date: July 25, 2000

Title: "The Dynamic Synapse Speech Recognition System: Using Neural Nonlinear Dynamics for Temporal Pattern Recognition"

Speaker(s): Dr. Theodore W. Berger & Dr. Jim-Shih Liaw

Affiliation(s): Department of Biomedical Engineering, Center for Neural Engineering, University of Southern California

Abstract:
Drs. Berger and Liaw will describe a novel neural network architecture based on the nonlinear dynamics of synaptic transmission in the hippocampus, a part of the brain involved in the formation of pattern recognition memories. A combined experimental-theoretical approach based on nonlinear systems theory is used to characterize functional properties of hippocampal neurons and synapses, and in particular, those properties that underlie the sensitivity of hippocampal neural elements to higher-order temporal patterns. These nonlinear transformational characteristics are then embedded in neural network models, and used as the instruments to extract features of temporally coded inputs, e.g., speech signals. A novel "dynamic learning rule," based on adaptive mechanisms of hippocampal synapses, is used to obtain an optimized feature set. Results demonstrate that this approach provides the basis for speaker-independent and speaker-specific word recognition with very small, highly simplified neural networks. Performance of trained networks is highly robust with respect to noise, with systems to date out-performing both human listeners and commercial speech recognition systems, when either Gaussian white noise or background speech from multiple, non-target speakers is used. Because the model assumes only neurobiological properties, the system also can be extended to other application domains involving temporal or spatio-temporal pattern recognition, e.g., speaker verification/identification, sonar classification, and sensor fusion. Also to be reviewed is related work involving analog VLSI implementations of the nonlinear neural network models, including those developed for speech recognition.

RIACS Seminar #34

Date: July 24, 2000

Title: "The Evolutionary Intelligent Agent Concept"

Speaker(s): Dr. Paul Dan Cristea

Affiliation(s): Professor, Department of Engineering Sciences, "Politehnica" University of Bucharest

Abstract:
The presentation will focus on preliminary results in exploring the Evolutionary Intelligent Agent (EIA) concept, that merges the Intelligent Agent and the Evolutionary - Genetic Algorithms approaches. EIAs are Intelligent Agents provided with a genotype that controls their capabilities to carry out various tasks, i.e., their phenotype. EIAs benefit of the two major forces of adaptation: learning - occurring at the level of each agent and at the time scale of an agent's life, and evolution - taking place at the level of the population and unfolding at the time scale of successive generations. Thus, EIAs can address the real-life problem of adaptation in complex and non-predictable environments as the nowadays worldwide computer networks. In this phase, a prototype of the EIA system has been implemented for study purposes, to experimentally investigate the EIA concept. The model is quite simple, but illustrates the basic features of an EIA system. According to the concept, the EIAs have not only a reactive behavior, but also cognitive features. To help visualize the EIA system dynamics, a sensorimotor type of agents has been considered, evolving in a two-dimensional world and performing several simple tasks. The system can comprise one or more agent populations (teams). The agents from different populations interact only by acting in the same environment. The agents from the same population may also interact directly, e.g., through message exchanges, genetic interactions, etc. An agent holds subjective, partial information about the environment, at two levels of world representation: the sensorial level - depicted in a sensorial map constructed with the tactile and visual inputs, and the cognitive level - depicted in a cognitive map based on the information in the sensorial map, but modified and enriched through some heuristic processing and with the information received by communicating with other agents in the same population.

RIACS Seminar #33

Date: July 24, 2000

Title: "Self-Interested Agents: From Competition to Cooperation"

Speaker(s): Dr. Adina Magda Florea

Affiliation(s): Professor, Department of Engineering Sciences, "Politehnica" University of Bucharest

Abstract:
In a multi-agent system, an agent exists and performs its activity in a society in which other agents exit and act. Therefore, coordination among agents is essential for achieving the goals and acting in a coherent manner. When agents in the system are self-interested, coordination and cooperation should be achieved through communication and negotiation. The presentation will focus on two aspects of self-interested cognitive agents behaviour. A first part is dedicated to present a model and associated behaviour of self-interested agents that are endowed, besides the widely accepted BDI attitudes, with preferences, obligations, norms, and gain. The behaviour of the agents is mainly motivated by the gain they obtain while fulfilling their preferred goals and by the necessity to cooperate with other agents for achieving these goals. The main focus is on negotiation, an agent using a set of inference rules based on cost, gain, and the cooperation profile of the other agents developed during previous interactions. The model is fitted to develop multi-agent based applications for open environments but also for applications with cooperative agents in which agents are distributed and have a certain degree of independence. A second part of the presentation is dedicated to a multi-agent system that tries to solve the problem of rational exploitation of natural renewable resources, named also the "tragedy of commons" problem. The agents use a genetic representation to model the evolution of the unpredictable world in which they live, leading to ecological plans of actions to preserve the resources. The genetic approach is based on cooperative coadapted species and models the multi-agent world from the point of view of a particular agent. The proposed genetic representation may also be used to investigate, in a centralized manner, the behavior and features of the entire society of agents for the "tragedy of commons" problem. The aim of the work is to propose and investigate a hybrid type of intelligent agents that are basically cognitive but which are endowed with evolutionary components to overcome the limitations of their beliefs and lack of complete knowledge on the environment in which they act.

RIACS Seminar #32

Date: July 11, 2000

Title: "Intelligent Systems - USC/ISI"

Speaker(s): Yigal Arens

Affiliation(s): USC/ISI

Abstract:
The Information Sciences Institute of the University of Southern California is a computer science research facility located off-campus in Marina del Rey. It is active in a number of areas of computer science, including networking, software engineering, and artificial intelligence. The Intelligent Systems Division at ISI is one of the largest AI groups in the country, with over 70 staff researchers, research faculty, students and visitors. AI research at ISI includes research in the areas of knowledge representation and ontologies, information management and integration, intelligent agents, natural language translation and summarization, modeling and simulation, distance education, digital government, robotics, and more. My talk will describe the main AI projects at ISI, devoting special attention to those that overlap with ongoing research at RIACS.

RIACS Seminar #31

Date: July 7, 2000

Title: "Statistical Machine Learning for Large Scale Optimization"

Speaker(s): Dr. Arun Jagota

Affiliation(s): U.C. Santa Cruz

Abstract:
Difficult optimization problems arise in all fields of science, engineering, and industry; exact solution methods are often lacking, heuristic methods of all sorts abound. Recently, a group of researchers has begun to add yet another such meta-method to the mix. Optimization involves a search for good solutions. During the course of a search one visits several feasible/infeasible/partial/complete solutions. Perhaps something can be learnt from them, to be profitably applied to discovering a better solution, or perhaps towards solving a related problem. In this talk I will present an overview of the recent works of several groups on applying machine learning methods to optimization problems. I will also discuss my own work in this area.

Please refer to:
http://www.cse.ucsc.edu/~jagota/csli-talk (containing a draft survey).

jagota@cse.ucsc.edu

RIACS Seminar #30

Date: July 6, 2000

Title: "Applying Aspect-Oriented Programming to the Intelligent Synthesis Environment Problem"

Speaker(s): Robert Filman

Affiliation(s): RIACS, NASA Ames Research Center

Abstract:
A critical issue in developing component-based and distributed systems is getting the assembled set of components to follow the policies of the overall system. To achieve ilities such as reliability, availability, responsiveness, performance, security, and manageability, all system components must consistently perform certain actions. One emerging example of a system in need of such regularity is NASA's effort to create an Intelligent Synthesis Environment (ISE).ISE aims to link scientists, design teams, manufacturers, suppliers and consultants in the virtual creation and operation of aerospace systems. In this talk, I describe the ISE problem and architectural issues for ISE.

A solution to the architecture complexity of ISE lies in employing the technology of Aspect-Oriented Programming (AOP). AOP provides mechanisms for separating the specification of different ilities and weaving them together into running systems. I will present an AOP system we have implemented, the Object Infrastructure Framework (OIF), and discuss how OIF could be applied to the ISE problem.

RIACS Seminar #29

Date: June 28, 2000

Title: "Overview of Research at ICASE"

Speaker(s): Manny Salas

Affiliation(s): ICASE, NASA Langley Research Center

Abstract:
The Institute for Computer Applications in Science and Engineering (ICASE) was established at NASA Langley Research Center on July 1972. The purpose of the institute when it was created was not only to do unclassified basic research in applied and numerical mathematics, fluid mechanics and applied computer science, but also to infuse the results of such research into Langley's programs, to expose the academic community to problems and programs of interest to NASA, an to foster interactions with that community. While research areas have changed over the years, the mission of the institute has remained true.

For 28 years ICASE has played a critical role in advancing the state of the art in computational fluid dynamics, turbulence modeling, boundary layer stability and transition, parallel numerical algorithms, and parallel compilers. Today, in addition to these research areas, ICASE conducts research in structures and materials, formal methods, multidisciplinary optimization and other areas of interest to Langley.

In this lecture I will give an overview of the institute and of its current research topics.

Manuel D. Salas, Director
ICASE
NASA Langley Research Center
Mail Stop 132C
Hampton, VA 23681-2199

RIACS Seminar #28

Date: June 22, 2000

Title: "Logic-Based Subsumption Architecture"

Speaker(s): Pedrito Maynard-Reid II

Affiliation(s): Stanford University

Abstract:
We describe a logic-based AI architecture based on Brooks' subsumption architecture. In this architecture, we axiomatize different layers of control in First-Order Logic (FOL) and use independent theorem provers to derive each layer's outputs given its inputs. We implement the subsumption of lower layers by higher layers using circumscription to make assumptions in lower layers, and nonmonotonically retract them when higher layers draw new conclusions. We also give formal semantics to our approach. Finally, we describe four layers designed for the task of robot control and an experiment that empirically shows the feasibility of using fully expressive FOL theorem provers for robot control with our architecture.

http://www-formal.stanford.edu/eyal/papers/lsa-ijcai99.ps

RIACS Seminar #27

Date: June 8, 2000

Title: "Interactive Visualization of Large Graphs and Network"

Speaker(s): Tamara Munzner

Affiliation(s): Stanford University

Abstract:
Many real-world domains can be represented as large node-link graphs: for instance, backbone Internet routers connect with 70,000 other hosts, mid-sized Web servers handle between 20,000 and 200,000 hyperlinked documents, and dictionaries contain millions of words defined in terms of each other. Computational manipulation of such large graphs is common, but previous tools for graph visualization have been limited to datasets of a few thousand nodes. Visual depictions of graphs and networks are external representations that exploit human visual processing to reduce the cognitive load of many tasks that require understanding of global or local structure. We assert that the two key advantages of computer-based systems for information visualization over traditional paper-based visual exposition are interactivity and scalability. We also argue that designing visualization software by taking the characteristics of a target user's task domain into account can lead to systems that are either more effective and scale to larger datasets than previous work. This talk includes a analysis of three specialized systems for the interactive exploration of large graphs, relating the intended tasks to the spatial layout and visual encoding choices. We present two novel algorithms for specialized layout and drawing which use quite different visual metaphors. The H3 system for visualizing the hyperlink structures of web sites scales to datasets of over 100,000 nodes by using a carefully chosen spanning tree as the layout backbone, 3D hyperbolic geometry for a Focus+Context view, and provides a fluid interactive experience through guaranteed frame rate drawing. The Constellation system features a highly specialized 2D layout intended to spatially encode domain-specific information for computational linguists checking the plausibility of a large semantic network created from dictionaries. The Planet Multicast system for displaying the tunnel topology of the Internet's multicast backbone provides a literal 3D geographic layout of arcs on a globe to help MBone maintainers find potentially misconfigured long-distance tunnels. Each of these three systems provides a very different view of the graph structure, and we evaluate their efficacy for the intended task. We generalize these findings in our analysis of the importance of interactivity and specialization for graph visualization systems that are effective and scalable.

Speaker's Bio:
Tamara Munzner is a PhD candidate at Stanford University, where she received a BS in computer science in 1991. In the intervening years she was a member of the technical staff at the Geometry Center, a mathematical visualization research group at the University of Minnesota. She was one of the authors of Geomview, the Center's public domain interactive 3D visualization system. While at the Center she was co-director and one of the animators of two computer generated mathematical videos, Outside In and The Shape of Space. Her current research interest is information visualization, specifically interactive techniques for visualizing large graphs and networks. Her recent research work on interactive layout and drawing of large graphs in 3D hyperbolic space has been commercialized through Silicon Graphics in the Site Manager application for webmasters.

RIACS Seminar #26

Date: June 6, 2000

Title: "Seminar on Distributed Shared-Memory Programming with the Unified Parallel C Information Power Grid Seminar"

Speaker(s): Dr. Tarek El-Ghazawi

Affiliation(s): George Mason University

Abstract:
Parallel programming paradigms have been designed around three models. These are message passing, data parallel, and shared- memory. Shared-memory can simplify programming, as it provides a memory view similar to that of uniprocessors. Practical experience, however, has shown that large-scale parallel machines should have physically distributed memories. Further, it has shown that as the programmer gets closer to the underlying hardware, higher performance execution could be achieved. Thus, designing parallel programming languages around a distributed shared-memory model has the promise of ease-of-programming as well as efficiency, since programmers can exploit features such as memory locality in distributed memory systems. Furthermore, the use of an abstract distributed shared-memory model can lead to program portability and allow efficient compiler implementation on other parallel architectures.

UPC is supported by a forum of government, academia, and industry, and many implementations are available or will become available soon.

Dr. El-Ghazawi will also present an overview of his previous research in high-performance computing and future direction.

RIACS Seminar #25

Date: May 11, 2000

Title: "Optimal Reward Functions in Distributed Reinforcement Learning"

Speaker(s): Dr. Kagan Tumer

Affiliation(s): NASA Ames Research Center, Code IC

Abstract:
The mathematics of ``COllective INtelligence'' (COINs) is concerned with the design of multi-agent systems in order to optimize an overall global utility function when those systems lack centralized communication and control. Typically in COINs each agent runs a distinct Reinforcement Learning (RL) algorithm so that much of the design problem reduces to how best to initialize/update each agent's private utility function, so as to avoid their working at cross purposes as far as the global utility is concerned. Traditional ``team game'' solutions to this problem assign to each agent the global utility as its private utility function. In previous work we used the COIN framework to derive the alternative ``Wonderful Life Utility" (WLU), and experimentally established that having the agents use it induces global utility performance up to orders of magnitude superior to that induced by use of the team game utility. In this work we discuss new utility functions that provide improvements over both the WLU and the traditional approaches. 

RIACS Seminar #24

Date: April 27, 2000

Title: "Bayesian Super-Resolved Surface Reconstruction From Images"

Speaker(s): Dr. Vadim Smelyanskiy

Affiliation(s): RIACS, NASA Ames Research Center

Abstract:
Bayesian inference has been used successfully for many problems where the aim is to infer the parameters of a model of interest. In this talk we formulate the three dimensional reconstruction problem as the problem of inferring the parameters of a surface model from image data, and show how Bayesian methods can be used to estimate the parameters of this model given the image data. Thus we recover the three dimensional description of the scene. This approach also gives great flexibility. We can specify the geometrical properties of the model to suit our purpose, and can also use different models for how the surface reflects the light incident upon it. In common with other Bayesian inference problems, the estimation methodology requires that we can simulate the data that would have been recorded for any values of the model parameters. In this application this means that if we have image data we must be able to render the surface model. However it also means that we can infer the parameters of a model whose resolution can be chosen irrespective of the resolution of the input images, and may be super-resolved. Also, once the model is inferred we can produce high resolution images of a surface from the view points that are different from those of low-resolution input images. We present results of the large scale inference of surface models from simulated aerial photographs for the case of super-resolution, where many surface elements project into a single pixel in the low-resolution images.

RIACS Seminar #23

Date: March 2, 2000

Title: "Compiling Language Models from a Linguistically Motivated Unification Grammar"

Speaker(s): Manny Rayner

Affiliation(s): RIACS, NASA Ames Research Center

Abstract:
Systems now exist which are able to compile unification grammars into language models that can be included in a speech recognizer, but it is so far unclear whether non-trivial linguistically principled grammars can be used for this purpose.  We describe a series of experiments which investigate the question empirically, by incrementally constructing a grammar and discovering what problems emerge when successively larger versions are compiled into finite state graph representations and used as language models for a medium-vocabulary recognition task.

RIACS Seminar #22

Date: February 17, 2000

Title: "The Sun's Magnetic Attractions"

Speaker(s): Lyndsay Fletcher

Affiliation(s): Lockheed Martin, Missiles and Space

Abstract:
The last three decades, since the advent of space-borne solar observatories, has been a period of rapid flourishing in our understanding of the outer atmosphere of the sun.  Observations in the ultraviolet to X-ray parts of the spectrum have shown clearly that the structure, the dynamics and the evolution for the solar atmosphere are dominated by its magnetic field, which is generated by a dynamo located at the interface between the radiative core and the convection zone.  The manifestations of the sun's magnetic activity include sunspots, solar flares, coronal mass ejections, and aspects of the solar wind. I will attempt to summarize our current understanding of the magnetic solar atmosphere, with the aid of many beautiful images from satellites such as Yohkoh, TRACE and SoHO.  I will also touch upon the implications of solar magnetic activity for life on our planet.

RIACS Seminar #21

Date: February 3, 2000

Title: "Knowledge-Based Abstraction, Visualization, and Exploration of Time-Oriented Data"

Speaker(s): Yuval Shahar

Affiliation(s): Stanford University Medical Center

Abstract:
I will describe a conceptual and computational architecture called Knowledge-based Navigation of Abstractions for Visualization and Explanation (KNAVE).  KNAVE is a domain-independent framework specific to the task of interpretation, summarization, visualization, explanation, and interactive exploration in a context-sensitive manner through time-oriented raw data and the multiple levels of higher-level, interval-based concepts that can be abstracted from these data.  The KNAVE domain-independent exploration operators are based on the relations defined in the knowledge-based temporal-abstraction problem-solving method, which is used to abstract the data, and thus can directly use the domain-specific knowledge base on which that method relies.  Thus, the domain-specific semantics are driving the domain-independent visualization and exploration processes.  By accessing the domain-specific temporal-abstraction knowledge base and the domain-specific time-oriented database, the KNAVE modules enable users to query for domain-specific temporal abstractions and to change the focus of the visualization, thus reusing for a different task the domain model that has been acquired from the domain experts.  Although I will focus on the methodology, I also will present a preliminary evaluation of the KNAVE prototype in a medical domain.  Our experiment incorporated seven users, a medical patient record, and three complex temporal queries, typical of guideline-based care, that the users were required to answer and/or explore.  The results of that preliminary experiment have been quite encouraging.  The KNAVE methodology has potentially broad implications for tasks such as planning, monitoring, explanation, and interactive data mining of time-oriented data.

Speaker's Bio:
Yuval Shahar received his B.Sc. and M.D. degrees from the Hebrew University in Jerusalem, Israel (1981).  Following a clinical internship and service as head of a regional clinic, Dr. Shahar served as the head of the Computer Section in the Israeli Defense Force Medical Corps from 1985 to 1988.  Dr. Shahar pursued M.Sc. studies in theoretical computer science at Bar Ilan University, Ramat Gan, Israel (1988), M.Sc. studies in artificial intelligence at the computer science department of Yale University, New Haven, USA (1990), and a Ph.D. in Medical Information Sciences from Stanford University, CA, USA (1994).  Dr. Shahar is an Assistant Professor of Medicine and Computer Science at Stanford University, and a member of its Medical Informatics Program, in which he heads the planning and temporal reasoning group.  Dr. Shahar's main research interests include temporal reasoning, planning, information visualization, and medical decision analysis.

RIACS Seminar #20

Date: January 20, 2000

Title: "Modeling and Simulating Work Practices on the Moon - An Agent-Based Approach"

Speaker(s): Maarten Sierhuis

Affiliation(s): RIACS, NASA Ames Research Center

Abstract:
In recent years, interest in collaborative agents has increased due to the fact that most applications also require collaboration with other systems and their users. Although we have started from a different need, namely understanding the way people work, collaborate, and communicate in their environment, we have developed an agent-based simulation environment (Brahms) that deals with a lot of the same issues as the intelligent agent community is now addressing.

In this talk I will present a multi-agent model of the collaborative work practices of the Apollo astronauts during the Apollo Lunar Surface Experiments Package (ALSEP) offload from the Lunar Module. The goal of this experiment was to investigate the use of the Brahms-language in describing an existing, though mostly forgotten, work practice. The challenge we faced in this experiment was to investigate if our theory of modeling work practice, as implemented in the language, would be sufficient to describe the work practice in the chosen domain. I will present an overview of the Brahms language and environment, based on this Apollo example.

RIACS Seminar #19

Date: December 13, 1999

Title: "An Overview of the Spoken Language Translator"

Speaker(s): Manny Rayner

Affiliation(s): RIACS, NASA Ames Research Center

Abstract:
In this talk, I will give an overview of the Spoken Language Translator (SLT) project. SLT, which ran from 1992 to 1999 under sponsorship from Telia Research, Stockholm, was one of the first serious projects in the area of automatic translation of speech. The final prototype performed translation between English, French and Swedish in the domain of air travel inquiry systems, using a vocabulary of about 1500 words, and with an accuracy of about 75%.

Average processing times are around 4 seconds per utterance.

The SLT system consists mainly of general purpose domain-independent components which are tuned to the domain using supervised training techniques, and uses a hybrid architecture which combines deep and shallow processing methods. Many of the techniques developed under the project are applicable to other types of spoken language understanding task.

A book about the SLT system will be published next year (2000) by Cambridge University Press.

RIACS Seminar #18

Date: December 2, 1999

Title: "Combining Prosodic and Language Models for Speech Segmentation and Recognition"

Speaker(s): Andreas Stolcke

Affiliation(s): SRI International

Abstract:
This talk will give an overview of recent work at SRI aimed at extending the output of speech recognizers beyond the usual stream of words, providing additional information such as sentence segmentation, disfluency detection, identification of proper names and topic segmentation.  All of these tasks are addressed in a common approach: a combination of the (fairly standard) lexical modeling, using hidden Markov and statistical language modeling, and the (not-so-standard) exploitation of prosodic cues to the "hidden'' events (tags) sought.  I will discuss the modeling approaches used, and present results for sentence segmentation and disfluency annotation on the Switchboard corpus, as well as for sentence and topic segmentation in the Broadcast News domain.  Prosodic modeling is shown to improve performance on these tasks, in some cases dramatically. It turns out that hidden event modeling can also improve word recognition itself, by constraining hypotheses to be consistent with prosodic characteristics of hidden events.

This is joint work with Liz Shriberg, Dilek Hakkani-Tur and Gokhan Tur.

Bio: Andreas Stolcke is a Sr. Research Engineer in SRI's Speech Technology and Research Laboratory, where he has been working on statistical language modeling, spontaneous speech modeling, and large-vocabulary recognition since 1994.  Prior to that, he received his undergraduate education at the Technical University of Munich, and a Ph.D. from UC Berkeley, with a thesis on Bayesian unsupervised learning of grammatical structure.

RIACS Seminar #17

Date: November 18, 1999

Title: "The Intelligent Systems Program"

Speaker(s): Butler Hine

Affiliation(s): NASA Ames Research Center

Speaker's Bio:
Dr. Butler Hine has just returned to NASA Ames as a new Level 2 Program Manager for the Intelligent Systems (IS) Program.  His area of responsibility will be the Autonomous Reasoning (AR) area, the goal of which is to create smart spacecraft, rovers, and ground systems capable of achieving mission goals independent of low-level human control.  Prior to returning to NASA Ames, Butler was CEO of a Silicon Valley software start-up company and before that was the head of the Intelligent Mechanisms Group in Code IC.  Butler's talk will begin with a brief summary of his background followed by a discussion of the goals and organization of the Autonomous Reasoning Program within IS. 

RIACS Seminar #16

Date: November 4, 1999

Title: "Model Checking for Autonomy Software"

Speaker(s): Charles Pecheur

Affiliation(s): RIACS, NASA Ames Research Center

Abstract:
To fulfill its mission of deep space exploration in a "faster, better, and cheaper" way, NASA is putting a lot of efforts in autonomous software such as Ames' Remote Agent (RA), demonstrated last May on Deep Space One (DS-1).  The validation of such complex software systems poses a big challenge: because of internal decision taking and concurrency, the range of possible situations becomes so large that traditional black-box testing is very inefficient.  Achieving and assessing the reliability of autonomous  system will require the use of advanced V&V techniques such as model checking.

After a succinct introduction to the ins and outs of model checking, this presentation will survey past, ongoing and planned work on applying model checking to autonomy software in NASA Ame's Automated Software Engineering Group.  As an initial case study, our team found important concurrency bugs in the executive subsystem of the RA, using the Spin model checker.  Building on initial work at CMU, we are now extending a compiler that feeds models used by the Livingstone fault recovery system into the SMV symbolic model checker.  Some similarwork has been performed on a model for the HSTS planner (HSTS and Livingstone are also part of the RA).  The next stage will be to address the validation of a model-based system as a whole, which poses a difficult problem: such a system is built around an inference engine that would be very hard to model in a form amenable to a model checker.

RIACS Seminar #15

Date: October 21, 1999

Title: "Turning Speech into Scripts"

Speaker(s): Manny Rayner

Affiliation(s): RIACS, NASA Ames Research Center

Abstract:
We describe an architecture for implementing spoken natural language dialogue interfaces to semi-autonomous systems, in which the central idea is to transform the input speech signal through successive levels of representation corresponding roughly to linguistic knowledge, dialogue knowledge, and domain knowledge. The final representation is an executable program in a simple scripting language equivalent to a subset of CSHELL.

At each stage of the translation process, an input is transformed into an output, producing as a byproduct a "meta-output" which describes the nature of the transformation performed. We show how consistent use of the output/meta-output distinction permits a simple and perspicuous treatment of apparently diverse topics including resolution of pronouns, correction of user misconceptions, and optimization of scripts. The methods described have been concretely realized in a prototype speech interface to a simulation of the Personal Satellite Assistant.

RIACS Seminar #14

Date: October 7, 1999

Title: "Barriers to Digital Libraries"

Speaker(s): Hector Garcia-Molina

Affiliation(s): Stanford University

Abstract:
In this talk, the speaker will discuss the challenges and barriers faced in implementing a modern digital library, including technical, legal, and societal ones. In particular, he will also give an overview of the Stanford Digital Library Project, and describe how we are addressing some of the technical barriers in the areas of interoperability, preservation, searching, mobile access, and protection of intellectual property.

RIACS Seminar #13

Date: September 23, 1999

Title: "A Transient Coupled Fluid/Solid Numerical Simulation Of The Aortic Valve For Normal And Calcified Leaflets"

Speaker(s): Dr. Andreas Frank

Affiliation(s): Stanford University

Abstract:
Aortic stenosis (AS) is an aortic valve disease causing flow obstruction during left ventricular (LV) systolic ejection due to a reduced aortic valve area (AVA). AS can be caused by calcification which limits leaflet excursion resulting in decreased AVA, increased flow resistance, and elevated LV pressures. Clinical assessment utilizes calculated AVAs, however, AVA may be a function of flow and thus not adequately discriminate between moderate and severe AS for varying flow conditions.

Numerical simulations for increased leaflet calcification (10x, 50x, and 200x normal) resulted in increased AS, as well as: 1) reduced leaflet displacement and AVA, 2) diminished slow closure phase, 3) increased transvalvular flow velocities with decreased flow, 4) increased transvalvular pressure gradients and resistance, 5) increased sinus vortex size, growth and strength, and 6) increased flow regurgitation. For severe AS with fixed calcification the resistance was nearly independent of flow, whereas the AVA was a linear function of flow (50% flow decrease provided 23% AVA decrease). Hence the resistance, not the AVA, better distinguished between moderate and severe AS for a variety of flow conditions, providing a better assessment of AS severity.

RIACS Seminar #12

Date: September 9, 1999

Title: "Mission Control: Technology Paradise or Technology Wasteland?"

Speaker(s): Dr. John O'Neill

Affiliation(s):

Abstract:
The Human-Centered Computing (HCC) group have been participating in the design of a new logging/handover tool to support the Reduced Increment Manning concept for the International Space Station's Mission Control. Reduced Increment Manning is exploring whether Mission Control can be operated without a full flight control team overnight and at weekends.

A key issue for ISS Mission Control is whether the artificial distinctions between "mission-critical" software and "PC-software" is meaningful. It can be argued that much of the information that has traditionally existed on the non-critical PC systems will actually be mission-critical in long duration missions.

The system integration challenges are more extensive than simply introducing new technologies. There are basic infrastructure issues that need to be addressed ranging from simple issues like networking the systems together through to more complex issues like how will ISS Mission Control at JSC collaborate with Mission Control at Moscow, with Mission Control at Marshall, and with Mission Control Centers for the Europeans, Japanese, and Canadians - and what technologies will enable these collaborations.

Finally, I will discuss some of the technological opportunities at Mission Control, and some of the paradigm-shifts that are preventing the introduction of new technologies.

RIACS Seminar #11

Date: August 25, 1999

Title: "Techniques for the Automated Restoration of Degraded Film and Video: Motion Estimation in the Presence of Missing Data"

Speaker(s): Dr. Anil Kokaram

Affiliation(s): Visitor to the Superresolution Group of Dr. Peter Cheeseman at RIACS

Abstract:
"The increasing importance of digital video in the broadcast television and film industries has placed heavy demand on holders of television and film archives. Unfortunately archived material is typically in a very bad condition. Scratches, missing data, line jitter, shake are all typical examples of degradation possible with time. It is expected that the consumer will come to demand a better quality image as part of the `digital package'. Furthermore, with the adoption of several compression schemes for broadcast, it is not sensible to waste bandwidth transmitting degraded portions of image material. Therefore removing these artefacts has been of interest for some time. Most of the post-production houses perform much of this re-touching by hand. This is a very painful process."

"This talk will introduce some of the problems in degraded video and film and present automated techniques for the treatment of the problems. Much of the material concerns the removal of missing data and an overview of a Bayesian approach to the problem will be given. In particular this solution has important implications for the estimation of motion in video sequences in the presence of missing data. An application for robust video communications in the presence of high error rates (e.g. wireless video, internet video) will also be illustrated.

"Equipment allowing, the talk will conclude with some video demos of restorations on real data."

Speaker's Bio:
Anil Kokaram completed the PhD at the Signal Processing Group of the Engineering Department, Cambridge University in 1993. He remained there as a research fellow till 1998 when he took up his present post of Lecturer in the Department of Electronic and Electrical Engineering at the University of Dublin, Trinity College, Dublin, Ireland. His main interest has been in the broad area of video processing and in problems in the restoration of degraded film and video including motion estimation for degraded sequences (e.g. Video acquired for particle image velocimetry). He has a particular interest in Bayesian techniques and MCMC for signal processing.

RIACS Seminar #10

Date: May 21, 1999

Title: "CMEX Experiments with the University of West Florida Institute for Human and Machine Cognition's (UWFa-IHMC) Concept Map Navigator for Educational Outreach and for Collaborative Research: a Precursor to More General Astrobiology Applications"

Speaker(s): Dr. Geoffrey Briggs

Affiliation(s): NASA Ames Research Center, Code S

Abstract:
"The Center for Mars Exploration (CMEX) is developing a new version of its teacher resource CD-ROM with a browser (developed by UWFa-IHMC) that uses concept maps to aid the user to better understand, and to more easily navigate within, the voluminous content of the CD. I will report on progress in the ongoing development of the CD in collaboration with the IHMC.

"CMEX is also preparing to use the Concept Map browser as part of its ongoing support for the Mars Surveyor Project Office in selecting landing sites for the upcoming lander/rover/sample return missions. Ames' CMEX is involved because astrobiology objectives are central to the missions' rationale. We (Virginia Gulick and others) are organizing annual science workshops to review current Mars mapping and analysis by the science community and, in support of these workshops, have established a web site that we intend to be state-of-the art for collaborative research. To this end CMEX is in a collaboration with Code IN (Glenn Deardorff and others) and we will report progress in developing the collaborative web site. It is hoped that this site will be a prototype for collaborative research within the Astrobiology Institute and the astrobiology community in general. Jeff Scargle will comment on the potential."

Speaker's Bio:
Dr. Geoffrey Briggs, Scientific Director of NASA Ames' Center for Mars Exploration (CMEX), will speak about an application of concept maps for navigating voluminous Mars data collections. Prior to his current position, Dr. Briggs was Director of NASA's Solar System Exploration Division in the Office of Space Science and Applications at NASA HQ. He was at HQ during the launch of the Pioneer Venus spacecraft, the Pioneer 11 encounter with Saturn, the launch of Galileo, Ulysses and Magellan and the Voyager launches, and the start of Magellan, Mars Observer and Cassini missions. Before serving at NASA HQ, Dr. Briggs worked at JPL and was a science investigator on the Mariner 9, Viking Orbiter and Voyager Imaging Teams.

RIACS Seminar #9

Date: May 7, 1999

Title: "The New Millennium Remote Agent: To Boldly Go Where No AI System Has Gone Before"

Speaker(s): Dr. Peter Norvig & Dr. P. Pandurang Nayak

Affiliation(s): NASA Ames Research Center, Code IC & RIACS

Abstract:
"The New Millennium Remote Agent is an autonomous spacecraft control system being developed jointly by NASA Ames and JPL. It integrates constraint-based planning and scheduling, robust multi-threaded execution, model-based diagnosis and reconfiguration, and real-time monitoring and control. The Remote Agent will control Deep Space One (DS-1), the first of NASA's New Millennium missions launching in late 1998. As the first AI system to autonomously control an actual spacecraft, the Remote Agent will enable the establishment of a "virtual presence" in space through an armada of intelligent space probes that autonomously explore the nooks and crannies of the solar system. In this talk I will describe the Remote Agent architecture and its main components, with a special focus on the model-based diagnosis and reconfiguration system embodied in the Livingstone program. Livingstone is a kernel of a reactive model-based autonomous system that performs significant deduction within the reactive control loop".

"Based on the IJCAI-97 Invited Talk given jointly with Nicola Muscettola, Barney Pell, and Brian C. Williams."

RIACS Seminar #8

Date: April 30, 1999

Title: "RIACS Language Interfaces and Speech Technology Group (RIALIST)"

Speaker(s): Dr. Yuri Gawdiak & Dr. Beth Ann Hockey & Dr. Frankie James

Affiliation(s): NASA Ames Research Center - Code IC & RIACS & RIACS

Abstract:
"There are many situations in which a standard desktop computer interface is difficult, dangerous, or impossible to use. For example, if the user is performing a task that requires their complete visual attention, a GUI display can be distracting. Similarly, if the user is wearing gloves or needs to use their hands to perform an experiment, keyboard input is tricky, at best. In these cases, other interface technologies, especially speech, can alleviate the problems faced by the users.

"Speech interfaces are becoming increasingly more common in desktop computing, with the advent of affordable dictation software by companies such as Dragon Systems and IBM. Flexible command and control applications, which are of the greatest potential use to NASA, are not the focus of commercial development. Since these systems are not designed with NASA in mind, they are therefore (1) not designed for use in physically challenging environments, and (2) lack specific technical language models for space and aeronautics domains.

"In this talk, we will discuss areas of speech technology of particular interest to NASA, including areas of basic research that will extend the capabilities of speech applications for supporting future NASA projects. As part of this discussion, we will outline our speech interface development for two current NASA projects: the Surface Movement Advisor (SMA) and the Personal Satellite Assistant (PSA)."

Speaker's Bio:
Beth Ann Hockey and Dr. Frankie James, members of the RIACS core research project, will present their plans and talk about the current status of their work in cooperation with the Personal Satellite Assistant (PSA) and Surface Movement Advisor (SMA) projects. Later in the afternoon Dr. Hockey and Dr. James will host a demonstration at their RIACS office. The cooperative effort will be introduced by NASA scientist Yuri Gawdiak, Level 2 Manager for the Aviation Safety Program, lead PI for the PSA project and project manager for SMA.

RIACS Seminar #7

Date: March 26, 1999

Title: "Human-Centered Computing: Methods, Program, and Progress"

Speaker(s): Dr. Michael Shafto & Dr. William Clancey

Affiliation(s): NASA Ames Research Center - Chief of Human-Automation Integration Research Branch & Chief Scientist of Human Centered Computing Technical Area

Abstract:
"In this presentation I will show how human-centered computing has developed at Ames into a research program with a portfolio of new projects and cross-center partnerships. I will explain the key ideas of participatory design and the total systems perspective that frame HCC research, and contrast HCC with traditional human factors and software engineering. I will describe how we are using the Intelligent Systems program to set up cross-center partnerships and outline a "project maturity model" by which progress over the next five years may be expected. Examples will show how HCC research covers the gamut from facilitation in computer system projects to inventing new modeling languages."

Speaker's Bio:
Dr. Clancey offers an important approach to modeling and improving human productivity within organizations. Dr. Michael Shafto, Chief of Human-Automation Integration Research Branch, will introduce Dr. Clancey.

RIACS Seminar #6

Date: March 19, 1999

Title: "Communication and Coordination in Mission Control"

Speaker(s): Dr. Michael Shafto & Dr. Erik Vinkhuyzen

Affiliation(s): NASA Ames Research Center - Chief of Human-Automation Integration Research Branch & RIACS Scientist, Human Centered Computing

Abstract:
Dr. Vinkhuyzen will show a video gathered during his work with the Flight Dynamics Officers in the Mission Control Center at JSC. The analysis of the materials will focus on the singular communications technology used by the Flight Controllers in Mission Control, the "voice loop system." Of special interest is the way individuals have adapted their natural communication skills to the capabilities and limitations of the voice loop system.

RIACS Seminar #5

Date: March 5, 1999

Title: "Overview of IT Research at ARC: Challenges and Opportunities"

Speaker(s): Dr. Steven Zornetzer

Affiliation(s): NASA Ames Research Center, Director of Information Systems Office

Abstract:
Dr. Zornetzer's talk will have broad appeal because NASA Administrator Dan Goldin has designated the Ames Research Center as NASA's Center for Excellence in Information Technology. Most RIACS scientists work in Code I, for this reason also the talk will be of special interest for us at RIACS. This will be a good time to meet Dr. Zornetzer, learn more about his vision for information technology and ask questions.

RIACS Seminar #4

Date: February 26, 1999

Title: "Center for Bioinformatics and the Joint BC/NREN Telemedicine Experiment"

Speaker(s): Dr. Muriel Ross & Dr. Marjory Johnson & Dr. Alexander Twombly

Affiliation(s): NASA Ames Research Center - Manager of Bioinformatics Center (Code SL) & RIACS & RIACS

Abstract:
The Center for Bioinformatics is dedicated to developing advanced, high fidelity 3-D imaging and interactive virtual environment technologies for biomedical and scientific purposes. This talk will emphasize those efforts directed toward the biomedical community. 3-D imaging is carried out from serial sections of biological and medical tissues and organs, whether these are obtained from microscopy or from CT or MRI scans. A special effort is underway to produce similarly high fidelity images from echo and sonic data, since these are possible to obtain on space station and spacecraft, or on distant planets. The images are to be used for diagnostic purposes and to simulate, in virtual environment, surgical procedures for planning and training purposes in medical settings on Earth. The long-term goal is to ensure the health of astronauts as they probe deeper into space, so that these same technologies will eventually find use on spacecraft should unanticipated medical problems arise. A natural outcome of this research is telemedicine, in which the goal is to bring the clinic to the patient rather than the patient to the clinic. Communication to spacecraft for medical advice is, after all, but another example of reaching into a remote site via telecommunication.

RIACS Seminar #3

Date: February 19, 1999

Title: "SOFIA: NASA's Stratospheric Observatory for Infrared Astronomy"

Speaker(s): Dr. Sean Casey

Affiliation(s): USRA, NASA Ames Research Center - Sr. Scientist, SOFIA Project

Abstract:
"I plan to present an overview of the facility covering development interests for both the observatory and science instrument programs. The question and answer period should cover the application of information research concepts to SOFIA as a demonstration platform."

RIACS Seminar #2

Date: February 12, 1999

Title: "Integration of Information from Satellite Images"

Speaker(s): Dr. Peter Norvig & Dr. Peter Cheeseman

Affiliation(s): NASA Ames Research Center - Chief of Computation Sciences Division (IC) & RIACS

Abstract:
Dr. Norvig will outline IC Division projects. Dr. Cheeseman will discuss the integration of information from satellite images.

RIACS Seminar #1

Date: February 5, 1999

Title: "Astrobiology"

Speaker(s): Dr. David Morrison

Affiliation(s): NASA Ames Research Center - Director of the NASA Ames Space Directorate

Abstract:
Dr. Morrison's directorate embraces a literal universe of scientific areas of intellectual interest, including (just a sampler): origin and evolution of galaxies, stars and planetary systems, origin and evolution of life, and possibilities for humans to inhabit other worlds. Some of Dr. Morrison's enabling goals are to establish a virtual presence throughout the solar system and develop revolutionary technologies for missions.