The department will soon be conducting interviews for open faculty positions within out dept. We have scheduled visits/seminars for 3 candidates so far. As part of this process we would like for you to participate by meeting and talking with these candidates as well. Dr. Lynne E. Parker Oak Ridge National Laboratory Monday, March 11, 2002 1:30-2:30 p.m. Claxton Complex - Room 206 Meeting with Graduate Students: 3/11 at 4:30 in C233 - CS Library Meeting with Undergraduate Students: 3/12 at 9:30 in C233 - CS Library Distributed Embodied Intelligence: Advances in Multi-Robot Cooperation ABSTRACT: Distributed embodied intelligence has the potential to revolutionize challenging applications in the 21st century, including mining, surveillance and reconnaissance, hazardous waste cleanup, and assurance of infrastructure integrity. Research in distributed embodied intelligence focuses on artificial intelligence approaches to multi-agent systems that must physically interact with the world. These systems are particularly challenging when the embodied agents have no possibility for global control, a global system representation, shared global goals, or globally consistent knowledge. Examples of distributed embodied intelligence include multi-robot systems and distributed sensor networks. This seminar will cover some of my work to date in distributed embodied intelligence, which has focused primarily on cooperative multi-robot systems. I will describe a software architecture I have developed, called ALLIANCE, that facilitates robust, fault tolerant cooperative control. This software architecture allows team members to respond robustly and reliably to unexpected environmental changes and modifications in the robot team that may occur due to mechanical failure, the learning of new skills, or the addition or removal of robots from the team by human intervention. I will present a proof showing that this approach is guaranteed to enable successful completion of the group task under certain conditions. I will then describe portions of an extended version of ALLIANCE, called L-ALLIANCE (for Learning- ALLIANCE), which incorporates an automated dynamic parameter update mechanism. I will show that finding an optimal solution for the parameter update problem is NP-hard, leading to the need for a heuristic approach to the parameter update strategy. The resulting mechanism allows teams of mobile robots to learn from their previous experiences with other robots, enabling them to select their own actions more efficiently on subsequent trials when working with "familiar" robots on missions composed of independent tasks. These architectures have been successfully implemented in several proof-of- principle applications, including "mock" hazardous waste cleanup, box pushing, simulated coal mining, and military surveillance and reconnaissance. I will present several videos of these physical and simulated robot implementations. These approaches represent the current state of the art in fault tolerant cooperative control in multi-robot teams, as no other approach has been able to reach the level of robustness that is achievable using the ALLIANCE and L-ALLIANCE architectures. ------------------------------------------------------------------------ Micah Beck - Seminar 3/4 at 1:30 in C206 Title - Scalable Sharing of Storage in the Wide Area Meeting with Graduate Students: 3/4 at 4:30 in C233 - CS Library Meeting with Undergraduate Students: 3/5 at 9:30 in C233 - CS Library Scalable Sharing of Storage in the Wide Area Abstract: This talk will explore the concepts and mechanisms underlying Logistical Networking, a revolutionary architectural approach to communication that synthesizes elements of storage and wide area networking systems that are traditionally considered orthogonal. Logistical Networking is modeled on IP networking, and so its architecture is a stack with physical media and OS drivers at the bottom; an innovative layer that enables the scalable sharing of storage called the Internet Backplane Protocol (serving a function analogous to IP); and then the exNode, a tool for aggregating resources and enabling valuable end-to-end services such as reliability, high performance, and security (serving a function analogous to TCP). Researcher at the Logistical Computing and Internetworking Laboratory at the University of Tennessee have been pursuing the development of these mechanisms and their integration into higher level middleware and application-level tools in an effort to create a new architecture for scalable computing in the wide area based on the successful architectural approach of the Internet. This new architecture offers a new framework for advanced Internet applications of all kinds, from scientific Grid computing to collaborative work to multimedia content delivery. Micah Beck has been a contributor to research ranging from Parallel and Distributed Systems to Languages and Compilers to Advanced Internetworking and Content Distribution. He began his career doing research in distributed operating systems at Bell Laboratories and received his Ph.D. in Computer Science from Cornell University (1992) in the area of parallelizing compilers. He then joined the faculty of the Computer Science Department at the University of Tennessee, where he is currently a Research Associate Professor working in distributed high performance computing, networking and storage; he is also a Director of the Logistical Computing and Internetworking Laboratory. An active participant in the Internet2 project, he has since 1997 led their Distributed Storage Infrastructure project, defining an advanced Content Distribution model to enable edge processing. In 2000 he joined with other members of this project drawn from industry and academia to found Lokomo Systems and he currently serves as Chief Scientist of that company. ------------------------------------------------------------------------ Jing-Ru C. Cheng Department of Computer Science and Engineering The Pennsylvania State University Wednesday, March 6, 2002 1:30-2:30 p.m. Claxton Complex - Room 206 Parallel Particle Tracking Methods For Scientific Computing ABSTRACT: Particle tracking methods are a versatile computational technique central to the simulation of a wide range of scientific applications including: visualization, molecular dynamics, direct simulation Monte Carlo methods, and Eulerian-Lagrangian methods. We introduce a common framework, the "in- element" particle tracking method, based on the assumption that particle trajectories are computed by problem data localized to individual element. The goal of the software implementation is to design the software interface to be portable between different programming environments, while the component interface is lightweight and functional -- allowing for the easy incorporation by other parallel programming environments. This presentation introduces the parallel algorithms, presents a new parallel approach for the dynamic partitioning of particle-mesh computational systems, and demonstrates a new \emph{a posteriori} error estimator based on particle tracking methods for adaptive mesh refinement. Experimental results detail the performance of this parallel load balancing particle method for two- and three- dimensional particle-mesh test problems on an unstructured, adaptive mesh. ------------------------------------------------------------------------ Lynne Parker - Seminar 3/11 at 1:30 in C206 Title - to be announced Meeting with Graduate Students: 3/11 at 4:30 in C233 - CS Library Meeting with Undergraduate Students: 3/12 at 9:30 in C233 - CS Library ------------------------------------------------------------------------ Chris Long - Seminar 2/25 at 1:30 in C206 Title - quill: A Gesture Design Tool for Pen-based User Interfaces Meeting with Graduate Students: 2/25 at 4:30 in C233 - CS Library Meeting with Undergraduate Students: 2/26 at 9:30 in C233 - CS Library Abstract: This talk will discuss the motivation, design, and development of a tool for designing gestures for pen-based user interfaces. Pen-based computers are becoming more common, especially small devices such as the Palm Pilot. One common technique in pen-based interfaces is using gestures--marks that invoke commands. Gestures can be intuitive and faster than other methods of invoking commands. However, gestures are sometimes problematic, in part because they are hard to design. To help interface designers create better gestures, I designed and built quill, an intelligent gesture design tool. It warns designers about gestures that may be hard for the computer to recognize or may be confused by people, and gives advice on how to fix these problems. An evaluation of quill revealed that its advice was helpful for some but not all designers. More work is needed to improve the usefulness of the advice to more designers. ------------------------------------------------------------------------