More information at http://web.sa.sc.edu/cgc/events/game-design-workshop/

ACM Talk Speaker: Alex Wong - Validation Engineer at Intel, an alumnus of the University of South Carolina Time: November 17th at 6:00pm. Place: Room 2A14/16 of Swearingen There will be pizza! Please RSVP on facebook: https://www.facebook.com/events/1514532028863460/

Dissertation Defense Candidate: Yang Song Advisor: Dr. Jason O’Kane Date: Friday, November 13, 2015 Time: 1:00pm Location: Swearingen, 3A00 (Dean’s Conference Room) Abstract This dissertation first presents our geometric-­based planning algorithm for robots to describe and reason about uncertainty, termed the `"constrained geometric approximation approach" Previous work delivered computationally expensive methods to represent a robot's knowledge of its states and uncertainty explicitly, using an information state in the information space. In contrast, our approach is computationally inexpensive. Specifically, this approximation method uses simple geometric data structures which are suitable for the robots with extreme limitations in both sensing and computation. We provided simulations of sensor ­based navigation tasks, along with experimental results. We conclude that the robot can achieve similar success rate at a small fraction of the computational cost, by maintaining simple representations of the incomplete information. The dissertation also contributes two decentralized algorithms for the multi-­robot lattice pattern formation problem. Prior work solved such problems by constituting only particular lattice patterns. Our approaches enable robots to form arbitrary lattice patterns without global knowledge of the system. Both of the formation algorithms feature in representing the lattice as a directed graph, in which edges show the desired rigid body transformations between the robot and its neighbors. Another common feature of the two algorithms is that robots organize themselves using tree structures. The first formation algorithm enables each robot to perform distributed task assignment using only local information. The experimental results demonstrate that the algorithm executes efficiently and scales reasonably well as the number of robots increases. Moreover, the algorithm is stateless and is robust to system failures. We analyze the pitfalls of the proposed task-assignment-­based formation algorithm. According to the analysis, we have designed another formation algorithm that has new features including: The bounded execution time, a novel motion strategy that guarantees the network connectivity, and the improved final lattice formation quality. Furthermore, we have proved the correctness of the new algorithm and conducted groups of experiments to compare the timing performance and the formation qualities of both formation algorithms.

THESIS DEFENSE Candidate: Melissa Henderson Advisor: Dr. Jenay Beer Date: Friday, November 6, 2015 Time: 2:15pm Place: Swearingen, 3A75 Conference Room Abstract Industrial applications have not been extensively researched regarding human-robot interaction. This project investigated industrial robot acceptance in the context of one manufacturing site. Acceptance was examined in relation to the workforce and conditions of acceptance were identified using mixed methodology. Quantitative (surveys) and qualitative (interviews) data was used to measure and analyze the existing state of technology acceptance and site culture. Based on this exploratory study, it was found the manufacturing facility has a weak culture but would be generally accepting of industrial robots if the technology were easy to use and useful. The identified boundary conditions to acceptance included training, job satisfaction and the opportunity to work in teams. It is hypothesized there was hesitation towards acceptance for age groups > 42 years and with less technology experience or exposure.

Candidate: Arjang Fahim Advisor: Dr. Homayoun Valafar Date: Monday, November 9, 2015 Time: 10:00am Place: Swearingen, 3A75 Conference Room Abstract Proteins are often referred as working molecule of a cell, performing many structural, functional and regulatory processes. Revealing the function of proteins still remains a challenging problem. Advancement in genomics sequence projects produces large protein sequence repository, but due to technical difficulty and cost related to structure determination, the number of identified protein structure is far behind. Novel structures identification are particularly important for a number of reasons: they generate models of similar proteins for comparison; identify evolutionary relationships; further contribute to our understanding of protein function and mechanism; and allow for the fold of other family members to be inferred. Considering the evolutionary mechanisms responsible for the generation of new structures in proteins, it has been speculated that there may be a limited number of unique protein folds as few as ten thousand families. Currently, the Protein Data Bank consists of nearly 113,000 protein structures, but less than 1,500 families are represented, and almost no new fold families have been reported since 2008. Ideally, solved protein structures for new protein families would be used as templates for in silico structure prediction methods, and the results of both solved and predicted structures would in turn be used to infer function. However, such an approach requires new, efficient and cost-effective computational methods for target selection and structure determination. Traditional characterization of a protein structure by NMR spectroscopy is expensive and time consuming regardless of the structural novelty of the target protein. In an effort to expand the applicability of NMR spectroscopy, the community is continually focused on the development of new and economical approaches that enable the study of more challenging, or structurally novel proteins. While many advances have been made in this regard, very little attention has been made on reducing the cost of structural characterization of routine proteins. Probability Density Profile Analysis (PDPA) has been previously introduced to directly address the economies of structure determination of routine proteins and subsequently, identification of novel structures from minimal sets of NMR data. The latest version of PDPA (2D-PDPA) has been successful in identifying the structural homologue of an unknown protein within a library of 1000 decoy structures. In order to further expand the selectivity and sensitivity of PDPA, incorporation of additional data is necessary. However, current PDPA approach is limited by its computational requirements, and its expansion to include additional data will render it computationally infeasible. Here we propose a new method and developments that eliminate PDPA's computational limitations and allow inclusion of Residual Dipolar Coupling (RDC) data from multiple vector types in multiple alignment media. Additionally nD-PDPA will be used to refine an unknown protein to obtain closer structure to the native in terms of bb-rmsd.

Please join us, as women in Computing at USC welcomes guest speaker Jill Solovey! Jill is the Director of Enterprise Solution Development at Premier, Inc., a healthcare IT company focused on transforming healthcare with collaboration and technology. Jill will speak about her experiences in computing and software engineering. As always, pizza will be provided. Date: Monday Nov. 2nd Time: 6:30pm Where: Swearingen 3A75 Please RSVP via Facebook: https://www.facebook.com/events/860256300755808/

COLLOQUIUM Department of Computer Science and Engineering University of South Carolina Brian Williams, Los Alamos National Laboratory Date: October 30, 2015 Time: 1420-1510 (2:20-3:10pm) Place: Swearingen 2A18 Abstract Bayesian design of experiments is a methodology for incorporating prior information into the design phase of an experiment. Unfortunately, the typical Bayesian approach to designing experiments is both numerically and analytically intractable without additional assumptions or approximations. In this paper we discuss how Gaussian processes can be used to help alleviate the numerical issues associated with Bayesian design of experiments. We provide examples drawn from accelerated life testing and probabilistic calibration of model parameters and compare our results with large sample methods. Brian Williams has been a Technical Staff Member at the Los Alamos National Laboratory (LANL) since 2003. He is a member of the Statistical Sciences Group. Previously he was associate statistician at the RAND Corporation (2000-2003). He is contributing to the development and implementation of statistical methods for the design and analysis of computer experiments, focusing on the technical areas of sequential optimization, global sensitivity analysis, calibration,predictive maturity assessment and rare event inference for computer models. He is currently working on the development of technology for efficient Bayesian experimental design and implementation of uncertainty quantification methods supporting applications in nondestructive assay. Dr. Williams coauthored a 2003 book entitled The Design and Analysis of Computer Experiments with Thomas J. Santner and William I. Notz of The Ohio State University. In 2015, he was elected Fellow of the American Statistical Association for fundamental methodological contributions to the statistical design of experiments involving computer simulators and the analysis of data from such experiments including uncertainty quantification, for excellence in leadership of uncertainty quantification in critical federal programs, for excellence in collaborative research, and for service to the American Statistical Association. His research interests include experimental design, computer experiments, Bayesian inference, spatial statistics, statistical computing, and uncertainty quantification. He holds the Ph.D. in statistics from The Ohio State University.

http://www.sc.edu/ofsp/last_lecture_series.shtml

Dr. Farkas will be part of the panel at 5pm.

Abstract: There are lots of practical reasons why one might attach a tether to a mobile robot (providing power from off-board sources, high-speed communication to a base-station, etc.) but, since the tether constrains the motion of the robot, doing so makes the problem of moving the robot trickier than it would be otherwise. This talk will explore the motion planning problem for a planar robot connected via a cable to a fixed point in R^2. I'll describe how to visualize the configuration space manifold for such a robot, showing that it has regularity which can be used to produce a neat representation. This representation describes the manifold via (1) a discrete structure that characterizes the cable's position (2) an element within a single continuous cell. Further, when the tether has a constraint on its curvature, I'll show how Dubins’s theory of curves can be combined with work on planning with topological constraints to concisely represent the configuration space manifold, resulting in a data-structure that facilitates search for optimal paths. Bio: Dylan Shell is a computer scientist with broad interests. He's an Associate Professor in the Department of Computer Science and Engineering at Texas A&M University, where he runs a laboratory focused on robotics and artificial intelligence. His research group aims to synthesize and analyze complex, intelligent behavior in distributed systems that exploit their physical embedding to interact with the physical world in a variety of ways. He has published papers on multi-robot task allocation, robotics for emergency scenarios, biologically inspired multiple robot systems, multi-robot routing, estimation of group-level swarm properties, minimalist manipulation, rigid-body simulation and contact models, human-robot interaction, and robotic theatre. His work has been funded by DARPA and the NSF; and he has been the recipient of the Montague Teaching award, the George Bekey Service award, and the NSF Career.