Visitors at the IDeA Labs

Matthew Kirchner

April 5, 2022

Matthew R. Kirchner received his B.S. in Mechanical Engineering from Washington State University in 2007 and his M.S. in Electrical Engineering from the University of Colorado at Boulder in 2013. In 2007 he joined the Naval Air Warfare Center Weapons Division in the Navigation and Weapons Concepts Develop Branch and in 2012 transferred into the Physics and Computational Sciences Division in the Research and Intelligence Department, Code D5J1000. He is currently a Ph.D. candidate in the Electrical and Computer Engineering Department at the University of California, Santa Barbara. His research interests include level set methods for optimal control, differential games, and reachability; multi-vehicle robotics; nonparametric signal and image processing; and navigation and flight control. He was the recipient of a Naval Air Warfare Center Weapons Division Graduate Academic Fellowship from 2010 to 2012; in 2011 was named a Paul Harris Fellow by Rotary International and in 2021 was awarded a Robertson Fellowship from the University of California in recognition of an outstanding academic record. Matthew is a student member of the IEEE.


Sam Payne

March 8, 2022

Sam Payne is a bioinformatics professor at BYU, with a focus on cancer, proteomics and genomics. He is PI of a cancer data analysis center through the National Cancer Institute in collaboration with NYU and Washington University (WUSTL). Prior work in algorithms for proteomics and genomics has been funded by the Department of Energy (Early Career Investigator) and the National Science Foundation.

Before joining BYU, Dr. Payne was an Senior Research Scientist at the Pacific Northwest National Laboratory and an Assistant Professor of Informatics at the J. Craig Venter Institute. Dr. Payne received a B.S. of Computer Science at Brigham Young University. He earned his Ph.D. in Bioinformatics from UC, San Diego working with Dr. Vineet Bafna.


Cem Aydin

March 3, 2022

Cem is a PhD student in the Management Science and Operations Department at London Business School. His dissertation is on the impact of sampling error on outcome-based payment schemes. These payment schemes are usually designed as large competitions between thousands of hospitals. Cem builds a finite player model and extends it to an infinite player model before using reduced form parametric reg to empirically find support for modeling results. He then uses stats and structural estimation ideas to approximate hospitals’ cost function and use that to do counterfactual policy experiments on how hospitals would behave if the government used a different type of estimator.


Keith Paarporn

January 16, 2022

Keith Paarporn received his B.S. in Electrical Engineering from the University of Maryland, College Park in 2013, his M.S. in Electrical and Computer Engineering from the Georgia Institute of Technology in 2016, and his Ph.D. in Electrical and Computer Engineering from the Georgia Institute of Technology in 2018. Since 2018, he is a postdoctoral scholar in the Electrical and Computer Engineering Department at the University of California, Santa Barbara. In 2020, he served as a research consultant for ArcTan, Inc. His research interests focus on the analysis and control of networked multi-agent systems.


Philip E. Paré

December 1, 2021

Philip Paré is an Assistant Professor in the Elmore Family School of Electrical and Computer Engineering Purdue University. He is a member of the Center for Innovation in Control, Optimization, and Networks (ICON) and is affiliated with the Integrative Data Science Initiative (IDSI) and the Center for Education and Research in Information Assurance and Security (CERIAS). Philip is also a member of the PIECE (Project for Inclusion in ECE) Committee.


Enoch Yeung

November 10, 2021

Enoch Yeung is an assistant professor of mechanical engineering at University of California, Santa Barbara he received a B.S. in Mathematics from Brigham Young University, Magna Cum Laude with University Honors, and a Ph.D in Control and Dynamical Systems from the California Institute of Technology.

His research interests are centered in system identification and machine learning, design and analysis of distributed control systems, and model reduction for stochastic and nonlinear systems, with applications to biological and cyber-physical systems.. His most recent work has focused on reverse-engineering genetic context effects in synthetic biocircuits, developing algorithms for network verification of synthetic biological networks, developing synthetic biosensors in living prokaryotes for performing temporal logic, and paper-based synthetic biosensing solutions for mobile devices. His latest work was showcased at the 2016 DARPA Wait, What? Technology Forum.

He is the recipient of an National Defense Science and Engineering Graduate Fellowship, National Science Foundation Graduate Fellowship, Kanel Foundation Fellowhsip, an ACC Best Presentation Session Award, and a Charles Lee Powell Foundation Fellowship. His research also has been supported by several research programs such as the DARPA Living Foundries program, the NSF Molecular Programming Project, and a AFOSR Biological Research Initiative.


Dan Broadbent

September 29, 2021

Dan Broadbent is the Physical and Computer Sciences Librarian at BYU HBLL, with subject specialties in animation, computer engineering, computer science, electrical engineering, information technology, and physics/astronomy.


Brock Kirwan

February 3, 2021

Brock Kirwarn is a professor of Psychology at BYU. He earned his BS in Psychology and Philosophy from University of Utah in 2001, his MA in Psychology from Johns Hopkins University in 2004, and his PhD in Psychological and Brain Sciences from Johns Hopkins University in 2006.

Dr. Kirwan studies the brain mechanisms that allow people to form and retain memories of events. This includes big events, like your wedding or your sixth birthday, as well as more mundane information, like where you parked your car or what the word “doughnut” means. One question that occupies much of Dr. Kirwan’s time is this: What information will we forget, and why? To address this and other questions, Dr. Kirwan and his students use a number of methods, including neuropsychological studies with memory-impaired patients, behavioral studies with healthy adults, and functional neuroimaging (fMRI) experiments. Their fMRI experiments are conducted at the new MRI Research Facility on BYU campus.


Mónika Józsa

February 27, 2017

Bio

Mónika Józsa received the M.S. degree in applied mathematics from Eotvos Lorand University in Budapest, Hungary in 2013. From 2013 she worked for Research Institute of Agricultural Economics in Budapest and from 2014 she started her Ph.D in systems and control at the University of Groningen. From 2016 she continued her Ph.D. at the University of Douai. Her research is on stochastic processes and system identification of stochastic systems.

Abstract

Granger causality defines prediction based causal relationship between stationary processes. We study the relationship between Granger-causality structure of a process and the graph structure of its state-space representations. We will show that Granger non-causalities between processes corresponds to zero-blocks of the system matrices in a linear time-invariant state-space representation of these processes. One of the applications is to model the causal structure of brain activity which will be illustrated with an example.


Enoch Yeung

January 26, 2017

Bio

Enoch Yeung received a B.S. in Mathematics from Brigham Young University, Magna Cum Laude with University Honors, and a Ph.D in Control and Dynamical Systems from the California Institute of Technology. His research interests are centered in system identification and machine learning, design and analysis of distributed control systems, and model reduction for stochastic and nonlinear systems, with applications to biological and cyber-physical systems.. His most recent work has focused on reverse-engineering genetic context effects in synthetic biocircuits, developing algorithms for network verification of synthetic biological networks, developing synthetic biosensors in living prokaryotes for performing temporal logic, and paper-based synthetic biosensing solutions for mobile devices. His latest work was showcased at the 2016 DARPA Wait, What? Technology Forum.

He is the recipient of an National Defense Science and Engineering Graduate Fellowship, National Science Foundation Graduate Fellowship, Kanel Foundation Fellowhsip, an ACC Best Presentation Session Award, and a Charles Lee Powell Foundation Fellowship. His research also has been supported by several research programs such as the DARPA Living Foundries program, the NSF Molecular Programming Project, and a AFOSR Biological Research Initiative.

Abstract

Over the past two decades, exponential advances in biotechnology and bioengineering have enabled us to explore different computing paradigms for biological systems. How do biological systems make decisions, grapple environmental complexity and uncertainty, and ultimately survive across non-stationary environments? In this talk, I will discuss several research thrusts that attempt to unravel some of the principles of nonlinear and stochastic computing in natural and engineered biological systems. First, I will introduce the concept of ‘biocircuit engineering’, or design of gene regulatory networks for specific computing tasks. I will then discuss how recent research in biocircuit verification methods have produced unexpected insight into a neglected form of gene control in synthetic gene networks, enforced by biophysical constraints. We demonstrate how these additional layer of computation can be used to greatly improve performance of a classical biocircuit, the toggle switch memory module. Our results reinforce the perspective that biological cells respond dynamically to their environmental context, to perform both local and global gene regulation. Moreover, our findings underscore a lack of modularity in biological computing that breaks away from traditional Von-Neumann architectures. Rather, we argue that cellular biological systems perform efficient computation using stochastic (and distributed) layered computational architectures (with feedback). These observations motivate ongoing efforts to develop a framework for engineering distributed layered computing architectures capable of processing and responding to data streams generated by stochastic and nonlinear systems.


Anurag Rai

January 11, 2017

Former IDeA Labs student Anurag Rai visited lab meeting and spoke about his current research at MIT in the Laboratory for Information and Decision Systems. His Ph.D. research focuses on backpressure routing algorithms yielding maximum throughput in networks. While at BYU, Anurag authored 6 papers and produced a masters thesis addressing the theory of structure in network systems.


Carolyn Beck

December 8, 2016

Bio

Carolyn Beck is currently an Associate Professor in the Department of Industrial and Enterprise Systems Engineering, at the University of Illinois at Urbana-Champaign. Her research activities are focused on the development of model reduction, clustering and aggregation methods, with applications in bioengineering and networks problems. Carolyn has been a visiting faculty at KTH in Stockholm (2013), Stanford University in California (2006) and Lund University in Lund, Sweden (1996). She has received national research awards (NSF CAREER and ONR Young Investigator), and local teaching awards.

Carolyn received her Ph.D. from Caltech, her M.S. from Carnegie Mellon, and her B.S. from California State Polytechnic University, all in Electrical Engineering. Prior to completing her graduate studies, she gained industry experience at Hewlett-Packard in Silicon Valley, where she worked as a Research and Development Engineer for 5 years.

Abstract

We consider the problem of clustering data sets where the data points are dynamic, or essentially time-varying. Our approach is to incorporate features of both the deterministic annealing algorithm as well as control theoretic methods in our computational solution. Extensions of our method can be made to the problem of aggregating time-varying graphs, for which we have developed a quantitative measure of dissimilarity that allows us to compare directed graphs of differing sizes. In this talk, an overview of our dynamic clustering algorithm will be given, along with some analysis of the algorithm properties. We will conclude with a few highlighted applications, and further extensions as time allows.


Dhruva V. Raman

March 3, 2016

Bio

Dhruva is a DPhil (i.e. PhD) student in the Control Group at the University of Oxford, under the supervision of Antonis Papachristodoulou and James Anderson. He completed an MMath Degree at the University of Warwick (2012) graduating with first class honours. From 2012-present, he has been on the Systems Biology Doctoral Training Centre, University of Oxford, fully funded through an EPSRC scholarship. This involved a year of taught courses in the broad field of Systems Biology, before he began his DPhil proper in October 2013.

Abstract

The local sensitivity of model predictions to parameter perturbation can be described as ‘sloppy’ when it is highly anisotropic with respect to perturbation direction. We extend the existing quantification of sloppiness to account for non-local perturbation, and find that the degree of sloppiness can be highly affected by perturbation length-scale. We then construct Hamiltonian flows tracing over parameter perturbations with a minimally disruptive effect on reference model predictions for each length scale, uncovering hidden conservation relations. Links to the concept of parametric unidentifiability are also provided.


Donatello Materassi

December 4, 2014

Bio

Donatello Materassi holds a Laurea in “Ingegneria informatica” and a “Dottorato di Ricerca” in Electrical Engineering/Nonlinear Dynamics and Complex Systems from Universita’ degli Studi di Firenze, Italy. he has been a research associate at the University of Minnesota (Twin Cities) till 2011, is a post-doctoral researcher at Laboratory for Information and Decision Systems (LIDS) at the Massachusetts Institute of Technology and a lecturer at Harvard University till 2014. That same year, he became a an assistant professor at University of Tennessee in Knoxville. His main research interests are graphical models, stochastic systems and cybernetics.

Abstract

The interest for networks and dynamical systems has been increasing in the past years, especially because of their capability of modeling and describing a large variety of phenomena and behaviors. The principal advantages provided by a networked system are a modular approach to design, the possibility of directly introducing redundancy and the realization of distributed and parallel algorithms. All these advantages have led to consider networked systems in the realization of many technological devices. At the same time, it is not surprising that natural and biological systems tend to exhibit strong modularity as well. Interconnected systems are successfully exploited to perform novel modeling approaches in many fields such as Economics, Biology, Cognitive Sciences, Ecology and Geology. While networks of dynamical systems have been deeply studied and analyzed in Physics and Engineering, there is a reduced number of results addressing the problem of reconstructing an unknown dynamical network, since it poses formidable theoretical and practical challenges.

One of the main challenges is definitely the identification of networked systems that can be extremely difficult to access or manipulate. This the necessity of general tools for the identification of networks that are known only via passive observation is rapidly emerging. The talk addressed this problem under several scenarios, trying to form a picture as general and complete as possible. A variety of techniques based on Wiener Filtering for the reconstruction of different classes of networks were introduced. Sufficient and necessary conditions for the correct detection of links were provided as well.


Randall Lewis

October 12, 2011

Bio

Dr. Randall Lewis is a 2006 graduate of Brigham Young University with degrees in both economics and mathematics. Dr. Lewis received his PhD from the Massachusetts Institute of Technology in 2010 with his fields of concentration being industrial organization and econometrics. While working toward his PhD, Dr. Lewis began working as an Economic Research Scientist for Yahoo! Research, where he is still currently employed. Dr. Lewis has been honored with various awards, including BYU’s Department of Economics Valedictorian and MIT Presidential Fellow in 2006 and Yahoo! Superstar Nominations in 2010 and 2011. At Yahoo! Research, Dr. Randall focuses on statistical measurement using econometrics and casual statistics to learn valuable insights from large data sets. He enjoys working on a variety of research fields ranging from the econometrics of social networks to large scale field experiments. Dr. Lewis has primarily focused his skills on measuring the impact of online display and search advertising on important business outcomes such as clicks, page views, searches, survey outcomes, and both online and offline (in-store) sales.

Abstract

Yahoo! Research partnered with a nationwide retailer to study the effectiveness of display advertising on online and in-store sales on more than three million shared customers. We measure the impact of higher ad impression frequency using a simple design that varies the ads that users see on the Yahoo! network within identically targeted campaigns: users in the treatment group see the retailer’s ads; users in the control group see unrelated, `control’ ads; and users in the half treatment group see an equal probability mixture of the retailer’s and control ads. We find a statistically significant increase in sales, relative to the control group, as a result of the ads. Doubling the number of impressions per person–from 17 to 34 in a two-week period–approximately doubled the treatment effect. We also find striking evidence that the ads most strongly affected customers who live closest to the retailer’s brick-and-mortar locations: those who live within two miles of a store experience more than ten times the incremental sales lift due to ads as those who live farther away.


Nghia Tran

June 19, 2011

Bio

Nghia Tran came to BYU in 2002 from a plantation in Vietnam. He joined IDeA Labs in 2004 and was involved in many projects such as market structure analysis and demand forecasting. The intimate professor-student working relationship provided by the research laboratories fostered an ideal setting for mentoring opportunities for Nghia. Utilizing such resources, Nghia was able to receive funding for his research in data mining in analysis of market power. He was awarded an ORCA Mentoring Grant in 2004 for this project.

Nghia graduated with a BS degree in Computer Science April, 2006 and continues his research in optimal path-reparametrization and market structure analysis. Nghia received grants from the BYU Office of Research and Creative Activities and won several other honors including 3rd place in the ACM Rocky Mountain Regional Programming Competition and honorable mention at the BYU Business Plan Competition. He defended his Masters thesis at BYU on August, 2009.


Sandip Roy

March 19, 2009

Bio

Dr. Sandip Roy received a B.S. degree in Electrical Engineering from the University of Illinois at Urbana-Champaign in 1998, and M.S. and Ph.D. degrees in Electrical Engineering from the Massachusetts Institute of Technology in 2000 and 2003, respectively. Since 2003, he has worked as an Assistant Professor at the Washington State University (WSU).
During his time at WSU, Dr. Roy has also held various outside summer appointments, including at the University of Wisconsin and NASA’s Ames Research Center. His research is focused on the control and design of complex dynamical networks, with application to air traffic control, sensor networking, and systems biology problems.

Abstract

The next time your flight from Salt Lake City is delayed, you may be astounded to hear that the following sequence of events led to the delay: 1) an adventurous squirrel shorted out a transformer in San Diego, causing a cascading power outage (and incidentally also starting a grass fire that closed a school); 2) the power outage impacted the computer system at Los Angeles airport, causing the airport to close for an hour (and also shutting down a grocery store’s refrigerator and so fomenting a salmonella outbreak); 3) the resulting backlog throughout the air traffic system caused your aircraft to be delayed. The increasing frequency of such bizarre events highlights that our infrastructures are becoming increasingly interconnected and stressed, and that this high connectivity is leading to increasingly far-reaching dynamical responses in the infrastructures.

In this talk, I contend that tools for shaping these global dynamics of infrastructure networks are badly needed, and introduce several strategies for shaping network dynamics using limited resources. Fundamentally, these various design strategies permit shaping of the network’s dynamics in the face of severe constraints and variations, by exploiting the interconnection (or graph) structure of the network.

Two infrastructure-network applications of our design strategies, in virus-spreading control and air traffic management, are pursued in the talk. Complementarily, allocation of scarce resources to shape dynamics is also valuable for numerous networked communication and computation processes; I also briefly introduce applications of our tools in these fields. Finally, I describe my group’s future plans, including some exciting undergraduate- and graduate- student projects that we are pursuing under the auspices of a course entitled “Network Structure, Dynamics, and Control”.


Matthew Maxwell

March 17, 2009

Bio

Matthew began research with IDeA Labs in November 2004. He initiated a research project with the U.S. Bureau of Reclamation involving the modeling and control of Piute Dam in the Sevier River Basin of Utah. Matthew used this research to complete his honors thesis requirement and qualify for honors graduation. Matthew was also involved with the IDeA Labs demand forecasting group and researching methods to forecast demand for the BYU Bookstore. His interests are system identification, feedback control, dynamic systems, demand forecasting, and game theory.
Matthew was a recipient of the prestigious Barry M. Goldwater Scholarship for 2005. He has also been the recipient of two BYU Office of Research and Creative Activities grants for research and development of an appropriate software framework for web-based hydrological data display (2004) and the system identification process for Piute Dam/Sevier River (2005). Matthew has also received other scholarships including the Thomas and Dorothy Leavey Scholarship, the National Instruments Scholarship, and many full/half tuition BYU scholarships. Matthew graduated from BYU in April 2006 with a B.S. in Computer Science. Following graduation, Matthew went to pursue his PhD at Cornell University studying Operations Research


Gürdal Arslan

March 12, 2008

Bio

Gürdal Arslan received his Ph.D. degree in electrical engineering from the University of Illinois at Urbana-Champaign, in 2001. From 2001 to 2004, he was an Assistant Researcher in the Department of Mechanical and Aerospace Engineering, University of California, Los Angeles. In August 2004, he joined the Electrical Engineering Department at the University of Hawaii, Manoa. His current research interests lie in the design of cooperative (multi-agent) systems using game theoretic methods. Recent applications of his research include autonomous resource allocation for mission planning, multi-sensor deployment, traffic management, and cooperative multi-user MIMO signaling in wireless communication systems. He is a member of the IEEE Control Systems Society and he received the National Science Foundation CAREER Award on “Cooperative Systems Design - Stochastic Games Approach” in May 2006.

Abstract

We will overview some of the recent developments in the design of cooperative (multi-agent) systems, defined as systems of interconnected autonomous agents optimizing their own local objectives yet accomplishing a global objective. Cooperative systems design is a recent research theme that received significant attention primarily due to interest in designing “smart” vehicles with intelligent and coordinated action capabilities to achieve a system-wide objective. Other applications include multi-vehicle search and target assignment for military mission planning, multi-sensor deployment for anti-submarine warfare, cooperative multi-user MIMO signaling in wireless communication systems, distributed optimization in VLSI routing, congestion management in transportation systems. There are two key issues in designing such systems:

  1. designing local objectives, i.e., telling the autonomous agents what to optimize, and
  2. designing negotiation algorithms, i.e., telling the autonomous agents how to optimize.

Recent research shows that game theory is the most natural framework to analyze and synthesize cooperative systems. We will review some of the core concepts and tools provided by game theory to address those key issues involved in designing cooperative systems.

More Information Here


Brandon Rohrer

November 15, 2007

Bio

Machines that think and move as if alive have fascinated Brandon since the advent of the Transformers. He has pursued this interest through mechanical engineering degrees at BYU (BS ‘97) and MIT (MS ‘99, PhD ‘02) and through his research in the Cybernetic Systems Integration Group at Sandia National Laboratories in Albuquerque, NM. Current research topics include high-performance prosthetic sockets, human neural interface technologies, and biomimetic machine learning.

Abstract

The problem of unsupervised learning in an unmodeled agent of an unmodeled environment is one of the hard problems in intelligent robotics, but it is a reasonable description of what human infants do. I will present a summary of some initial work I have done in this area–a Brain-Emulating Cognition and Control Architecture (BECCA).

A BECCA-driven agent bootstraps a model of itself and its environment through two simple algorithms: S-Learning and Context-Based Similarity. In S-Learning, sequences of experiences provide the basis for future predictions and command selection. Context-Based Similarity uses those sequences to form abstract concepts, dramatically reducing the dimensionality of the learning problem. Implementations of BECCA in simulation and in hardware will be given as illustrative examples.

More Information Here


Neil Dalchau

November 1, 2007

Bio

Neil Dalchau obtained an undergraduate masters degree in Mathematics (MMath) from the University of Oxford in 2005. During the final year, he also spent some time working on finite element methods for electromagnetic problems at Vector Fields Ltd. He is currently (2007) at the beginning of his third year at the University of Cambridge working in a collaborative project between Alex Webb (Dept. of Plant Sciences) and Jorge Goncalves (Dept. of Engineering). He is interested in Systems Identification techniques for linear and nonlinear systems, and network reconstruction methods. His research has used these tools to learn properties of uncharacterised mechanisms in the face of noisy biological observations. .

Abstract

Almost all organisms have evolved a circadian clock, a genetic network of interlocking feedback loops which provide temporal information at the cellular level. The circadian clock controls many physiological processes, conferring great advantages to the fitness of the organism. Circadian biology has seen great interest from mathematical modellers in recent years, due to the complex network of feedback loops. This work is predominately concerned with the core mechanism which generates the oscillations, the components of which are often known. In Arabidopsis thaliana, the model plant organism, many of the central oscillator genes are known, but the pathways through which they regulate physiology are often completely uncharacterised. We have been investigating experimentally and mathematically the interplay between the circadian clock and the uncharacterised signalling pathways of calcium (Ca2+), light and sugars.

In this talk, two studies will be presented. With a careful choice of training data, we show how linear systems can be used to predict Ca2+ dynamics resulting from uncharacterised pathways. These pathways are replaced with explicit time delays, and the resulting models show the necessity for two inputs controlling Ca2+ signals through extensive validation. Also, a bifurcation matching method is proposed for adjusting existing clock models to different experimental conditions (the external availability of sugars). This leads to testable hypotheses for the targets of sugar signalling in the circadian clock.


Peter Young

September 13, 2007

Bio

Dr. Peter M. Young received his Ph.D. in Electrical Engineering from California Institute of Technology in 1993, and worked for two years as a Postdoctoral Associate at Massachusetts Institute of Technology, before joining the faculty of Colorado State University in 1995. He has worked extensively on the development of advanced analysis and design techniques for large-scale uncertain MIMO systems, subject to both multiple uncertain parameters, and multiple dynamic uncertainties. This work provided a breakthrough in this area, where previous tools could only handle small problems because of the computational burden, and Dr. Young developed computational software packages which were released commercially as part of the MATLAB Robust Controls toolbox.

Currently Dr. Young is an Associate Professor at Colorado State University. His recent research interests include the development of analysis and design techniques for robust learning controllers, capable of adaptation whilst maintaining guaranteed robust stability. He has carried out this theoretical work in parallel with efforts in a number of specific application areas. These include structural vibration suppression and disk drive servo control, control of HVAC and energy storage systems, power system distribution grids and sustainable energy, and control of biological systems - specifically algae growth for biodiesel production.

Abstract

It is well known that real physical systems cannot be exactly described by mathematical models, though such models are a prerequisite for many controller analysis and design techniques. Thus one has to deal with the issue of “uncertainty” in mathematical models. Robust control theory deals with this issue by providing controllers which are robust to the uncertainty, i.e., they work for all allowed values of the unknowns. This provides rigorous stability and performance guarantees but necessarily sacrifices performance. The approach taken in learning control is to try to “learn” these uncertainties on-line in real-time, and hence converge to a controller that is specifically tuned to the dynamics of this particular plant. This has the potential to deliver “optimized” performance, but the problem arises as to how to get there. One typically has no a-priori guarantees about the performance or even stability of the learning controller. This is clearly not acceptable in a practical (non-simulation) environment. In this talk we will discuss a technique for the development of robust learning controllers. These attempt to deliver the best of both worlds, by using each approach to deal with the shortcomings of the other. Our application area for this work is HVAC systems. These are systems which are complex, time-varying, nonlinear, with poorly understood, but slow, dynamics. This makes them an ideal candidate for our approach, and we will discuss our results to date with an experimental testbed for HVAC control that we have developed.

More Information Here


Jorge Gonçalves

July 18, 2007

Jorge Gonçalves is a postdoctoral scholar in the Control and Dynamical Systems Divisions at Caltech.

Research Interests: Complex Systems, Modeling, analysis, and control of complex systems like biological metabolic networks, economic market, communication networks, statistical mechanics. Proof Methods. Hybrid Systems. Robustness analysis of nonlinear systems. Applications to robotic manipulators and walking robots.


Nicola Elia

April 6, 2006

Bio

Nicola Elia received the Laurea degree in Electrical Engineering from Politecnico of Turin in 1987, the Ph.D. degree in Electrical Engineering and Computer Science from Massachusetts Institute of Technology in 1996. He worked at Fiat Research Center from 1987 to 1990. He was Postdoctoral Associate at the Laboratory for Information and Decision Systems at MIT from 1996 to 1999. Presently is an associate professor with the Dept. of Electrical and Computer Engineering at Iowa State University. He received the NFS CAREER Award in 2001. His research interests include computational methods for controller design, complex systems, networked control systems.

Abstract

In this talk we consider systems involving communication channels in loops. In the first part, we focus on communication systems with access to feedback and show how such systems could be designed using control theory ideas and tools. We present a tight connection between the Sensitivity Bode Integral formula, (a fundamental limitation of feedback systems), and the achievable communication rate expressed in term of the average Directed Information of the channel. We apply our approach to several Gaussian channels and Gaussian networks and show that either achieves or improves on the available results. In the second part of the talk, we concentrate on feedback control systems over communication channels. We describe some new results on performance limitations induced by the presence of Gaussian channels in the feedback loop. We then consider fading channel models. The simplest model is the analog erasure channel, which is used as a model for packet-drop links. We present a general framework to analyze the performance of linear systems with linear controllers over fading channels. The approach is to consider the fading nature of the channels as a source of (stochastic) uncertainty, and to recast the whole problem as a robust control problem over stochastic perturbations. We present several examples to elucidate the setup and the framework, including the simultaneous design of controller encoders and decoders that exploit the channel state information. Finally, we apply the framework to predict the emer gence of power laws distributions in the behavior of networked control systems.

More Information Here


Robin Roundy

March 2, 2006

Bio

Robin Roundy is a professor of Operations Research and Industrial Engineering at Cornell University, where he has been since 1983. He graduated magna cum laude from Brigham Young University, where he received the Orson Pratt Award, which is given annually to the outstanding mathematics graduate. He then studied operations research at Stanford University, where he received his doctorate in 1984. That same year he won the Nicholson Student Paper Competition, sponsored by the Operations Research Society of America (ORSA). In 1985, he received a Presidential Young Investigator Award from the National Science Foundation. In 1988 he received the Fredrick W. Lanchester Prize of the Operations Research Society of America for the best paper of the year on operations research. Cornell’s College of Engineering awarded him the S. Yau Excellence in Teaching Award in 1997 and 2002. He is a member of The Institute for Operations Research and the Management Sciences, and of the Institute of Industrial Engineers.

Abstract

We summarize a multi-year research effort designed to provide useful tools for capacity planning decisions in the semiconductor industry. The decisions are crucial and challenging. The business environment is volatile, but equipment has long procurement lead times and is extremely expensive. Capacity planning in a stochastic environment. We will review and evaluate current business practices. We present methods for quantifying the errors in demand forecasts. We present a novel approach for multi-dimensional demand modeling, and discuss practical and algorithmic implications of different stock out cost models. We present efficient algorithms for provably solvable versions of the capacity planning problem.

More Information Here


David Robinson

November 13, 2003

Bio

David Robinson graduated with his bachelors in Mechanical Engineering from Brigham Young University in 1994. He immediately started graduate work, continuing with Mechanical Engineering at the Massachusetts Institute of Technology. His masters work dealt with the development of a rapid prototyping technology called 3D Printing, a layered manufacturing process. His doctoral research took place at the MIT Leg Laboratory. David’s dissertation focused on high fidelity, high power force controlled actuators. These actuators were used for dynamically stabilized walking and running robots developed in his research group.

At the completion of his doctoral program in 2000, David started work on a project, called “Ginger”, for the company that was to eventually become Segway LLC. After almost two years of strict confidentiality, David was relieved to finally show his family and friends that he had actually been doing something at work. David was responsible for system dynamics development of the Segway HT i-series, launched in December of 2001. More recently, David directed the system dynamics development and validation for Segway’s recently released second model, the Segway HT p-series. He is presently engaged in further core technology development for Segway.

David enjoys reading, swimming, biking, running, skiing and many other dynamically stabilized activities. He balances life too by staying actively involved with his family, church, and community. David is the proud father of three wonderful children and husband to a most incredible wife. David is a native of Provo, UT and is happy to be back home visiting his alma matter.

Abstract

The operation of the Segway HT has been described in the popular press as “magic.” But, the Segway HT is a real system and real systems have physical limits. The vast majority of engineering time in the development of a revolutionary product is spent in discovering, understanding, and dealing with those physical limits. In this presentation, I discuss a few of the major challenges in the development of the Segway HT with regard to the physical limits of the machine. I also talk about some of the practices fostered in our company culture that specifically address working at the limits.