Schedule of the Controls and Dynamic Systems (CDS) seminar series. Most of the seminars are held on Tuesdays from 4pm to 5pm (exceptions are highlighted) following a social hour from 3:30pm to 4:00pm.

Please click on each row for more information on the talk and speaker.

Onizaka conference room is located in the Aerospace wing of the Engineering Center (first floor).

Spring 2016

Speaker Institute Title Date Time Venue
Prof. Majid Zamani Department of Electrical and Computer Engineering, Technical University of Munich Automated synthesis of control systems: A double abstraction scheme Tuesday, May 10th, 2016 3:30pm-5:00pm Clark's Room (Dean's Office)

Abstract: Embedded control software plays a crucial role in many safety-critical applications: modern vehicles, for instance, use software to control steering, fuel injection, and airbag deployment. These applications are examples of cyber-physical systems (CPS), where software components interact tightly with physical systems. Although CPS have become ubiquitous in modern technology due to advances in computational devices, the development of core control software running on these systems is still ad hoc and error-prone. In this talk, I will propose a transformative design process, in which the controller code is automatically synthesized from higher-level correctness requirements. First, a compositional construction of abstractions of interconnected continuous systems is proposed. Those abstractions, themselves continuous systems, act as substitutes in the controller design process due to having possibly lower dimensions and simple interconnection topologies. Second, an automatic controller synthesis scheme is proposed by constructively deriving finite abstractions of infinite approximations of original continuous systems. The proposed automated synthesis of embedded control software holds the potential to develop complex yet reliable large-scale CPS while considerably reducing verification and validation costs.

Bio: Majid Zamani is an assistant professor in the Department of Electrical and Computer Engineering at Technical University of Munich where he leads the Hybrid Control Systems Group. He received a Ph.D. degree in Electrical Engineering and an MA degree in Mathematics both from University of California, Los Angeles in 2012, and an M.Sc. degree in Electrical Engineering from Sharif University of Technology in 2007. From September 2012 to December 2013, he was a postdoctoral researcher in the Delft Centre for Systems and Control at Delft University of Technology. Between December 2013 and May 2014, he was an assistant professor at Delft University of Technology.
Prof. Naomi Leonard Department of Mechanical and Aerospace Engineering, Princeton On the Nonlinear Dynamics of Collective Decision-Making in Nature and Design Thursday, April 28th, 2016 3:30pm-5:00pm Clark's Room (Dean's Office)

Abstract: The successful deployment of complex, multi-agent systems requires well-designed, agent-level control strategies that accommodate sensing, communication, and computational limitations on individual agents. Indeed, many applications demand system-level dynamics to be robust to disturbance and adaptive in the face of changes in the environment. Remarkably, animal groups, from bird flocks to fish schools, exhibit just such robust and adaptive behaviors, even as individual animals have their own limitations. To better understand and leverage the parallels between networks in nature and design, a principled examination of collective dynamics is warranted. I will describe an analytical framework based on nonlinear dynamical systems theory for the realization of collective decision-making that allows for the rigorous study of the mechanisms of observed collective animal behavior together with the design of distributed strategies for collective dynamics with provable performance.

Bio: Naomi Ehrich Leonard is the Edwin S. Wilsey Professor of Mechanical and Aerospace Engineering and an associated faculty member of the Program in Applied and Computational Mathematics at Princeton University. She is Director of Princeton's Council on Science and Technology and an affiliated faculty member of the Princeton Neuroscience Institute and Program on Quantitative and Computational Biology. Her research and teaching are in control and dynamical systems with current interests in coordinated control of multi-agent systems, mobile sensor networks, collective animal behavior, human decision dynamics, and intersections with dance. She is a MacArthur Fellow, a member of the American Academy of Arts and Sciences, and a Fellow of the IEEE, ASME, SIAM, and IFAC. She received the B.S.E. degree in Mechanical Engineering from Princeton University and the M.S. and Ph.D. degrees in Electrical Engineering from the University of Maryland.
Prof. Lucy Pao Department of Electrical, Computer, and Energy Engineering, CU-Boulder Combined Feedforward/Feedback Control of Flexible Structures: Recurring Themes across Diverse Applications Thursday, April 21th, 2016 3:30pm-5:00pm Koelbel 203 (Business School)

Abstract: In the past, robotic manipulators, machine tools, measurement devices, and other systems were designed with rigid structures and operated at relatively low speeds. With a growing demand for fuel efficiency, smaller actuators, and speed, lighter weight materials are increasingly used in many systems, making them more flexible. Achieving high-performance control of flexible structures is a difficult task, but one that is now critical to the success of many important applications, such as atomic force microscopes, disk drives, tape drives, robotic manipulators, gantry cranes, wind turbines, satellites, and the space station remote manipulator system. The unwanted vibration that results from maneuvering or controlling a flexible structure often dictates limiting factors in the performance of the system. Over the last few decades, many feedback, feedforward, and combined feedforward/feedback control methods have been developed for flexible structures. We will discuss and compare several of these control methods in conjunction with overviewing some of the issues in the modeling of flexible structures, and we will highlight a few recurring themes across the diverse application areas mentioned above.

This will be a very preliminary version of a plenary talk that I will be giving at the American Control Conference in July 2016, and I would welcome any comments or suggestions on the talk.

Bio: Lucy Pao is currently a Professor in the Electrical, Computer, and Energy Engineering Department and a Professor (by courtesy) in the Aerospace Engineering Sciences Department at the University of Colorado Boulder. She earned B.S., M.S., and Ph.D. degrees in Electrical Engineering from Stanford University. Her research has primarily been in the control systems area, with applications ranging from atomic force microscopy to disk drives to digital tape drives to megawatt wind turbines and wind farms. Selected recent honors include elevation to IEEE Fellow in 2012, the 2012 IEEE Control Systems Magazine Outstanding Paper Award (with K. Johnson), election to Fellow of the International Federation of Automatic Control (IFAC) in 2013, and the 2015 SIAM J. Control and Optimization Best Paper Prize (with J. Marden and H. P. Young). Selected recent and current professional society activities include being General Chair for the 2013 American Control Conference, an IEEE Control Systems Society (CSS) Distinguished Lecturer (2008-2014), a member of the IEEE CSS Board of Governors (2011-2013 (elected) and 2015 (appointed)), Fellow of the Renewable and Sustainable Energy Institute (2009-present), IEEE CSS Fellow Nominations Chair (2016- ), member of the IFAC Fellow Selection Committee (2014-2017), and member of the International Program Committees for the 2016 Indian Control Conference, the 2016 IFAC Symposium on Mechatronics Systems, and the 2017 IFAC World Congress.
Prof. Bahman Gharesifard Department of Mathematics and Statistics, Queen's University, Canada Structural controllability and stabilization of sparse bilinear control systems Thursday, April 14th, 2016 3:30pm-5:00pm Clark's Room (Dean's Office)

Abstract: In this talk, we present some results on the class of sparse bilinear control system, where the control and drift matrices are forced to be of certain sparse patterns. The class of sparse bilinear control systems models many practical scenarios of complex networked systems, for example chemical and microbial cell-growth.

In the first part, we study the structural controllability of the class of driftless sparse bilinear control system. We prove that sparse bilinear control systems are structurally controllable, hence extending the classical results of Boothby and Wilson in 1979 to the class of sparse bilinear control systems. Along the way, we prove that a driftless bilinear sparse control systems is controllable if and only if there exist two sparse matrices in the same class of sparse matrix system such that their corresponding bilinear control systems is controllable. Our result generalizes the results of Boothby in 1975 and moreover, our proof provides an alternative proof for the classical result of Kuranshi in 1951, which states that any real semi-simple Lie algebra can be generated by two elements of it. This part is a join work with Mohmmed Ali Belabbas (UIUC).

Time permitting, in the second part, using results from the theory of averaging of bilinear systems, we provide a graph theoretic characterization for the existence of periodic control inputs that stabilize a sparse matrix system to the origin. In particular, we introduce a class of extensions to the directed graph corresponding to a given sparse matrix system, which when contains a stable sparse matrix system implies that the original system is stabilizable using periodic inputs. When this condition holds, we provide a systematic procedure for designing such controllers.

The talk will be self-contained for most parts, and I present the technical and historical backgrounds necessary.

Bio: Bahman Gharesifard is an Assistant Professor with the Department of Mathematics and Statistics, Queen's University, Canada. Prior to joining Queen's, he was a Postdoctoral Research Associate with the Coordinated Science Laboratory (CSL) at the University of Illinois, Urbana-Champaign (2012-2013) and Postdoctoral Researcher at the Cymer Center for Controls and Dynamics at the University of California in San Diego (2009-2012). He received a PhD degree in Mathematics from Queen's University, Canada, in 2009. His research interests include systems and controls, distributed control and optimization, social and economic networks, game theory, geometric control and mechanics, and Riemannian geometry.
Prof. Dezhen Song Department of Computer Science and Engineering, Texas A&M Robotic Search of Transient Targets Thursday, April 7th, 2016 4:00pm-5:00pm Koelbel 203 (Business School)

Abstract: Searching for objects in physical space has been one of the most important tasks for mobile robots. Transient targets refer to intermittent signal emitting objects such as cellphone users, airplane black boxes, and unknown sensor networks. Searching for such targets is difficult because targets are found only if both signal emission and sensing range conditions are simultaneously satisfied. This problem is inherently stochastic which makes the traditional coverage-based searching techniques less effective. Considering a large searching region, sparse target distribution, the expected searching time, multi-target signal correspondence, variable signal transmission power, and the efficient coordination of multiple robots, we report a series of algorithms developed over last decade that handle the cases from single-target-single-robot to decentralized multi-target-multi-robot with different sensing and communication constraints and explicit performance analyses. Extensive simulation and physical experiment results are also included in the talk.

Bio: Dezhen Song is an Associate Professor with Department of Computer Science and Engineering, Texas A&M University, College Station, Texas, TX, USA. Song received his Ph.D. in 2004 from University of California, Berkeley, MS and BS from Zhejiang University in 1995 and 1998, respectively. Song's primary research area is perception, networked robots, visual navigation, automation, and stochastic modeling. Dr. Song received the Kayamori Best Paper Award of the 2005 IEEE International Conference on Robotics and Automation (with J. Yi and S. Ding). He received NSF Faculty Early Career Development (CAREER) Award in 2007. From 2008 to 2012, Song was an associate editor of IEEE Transactions on Robotics (T-RO). From 2010 to 2014, Song was an Associate Editor of IEEE Transactions on Automation Science and Engineering (T-ASE). Song is currently an Associate Editor for IEEE Robotics and Automation Letters (RA-L).
Prof. Rafael M. Frongillo Department of Computer Science, CU-Boulder Designing Adaptive Prediction Markets via Convex Geometry Thursday, March 31st, 2016 3:30pm-5:00pm Koelbel 203 (Business School)

Abstract: Prediction markets are a widely-used, accurate, engaging, and intuitive way to crowdsource probabilistic predictions of various outcomes, from basketball tournaments to political elections. Constructed as financial markets for securities whose payoffs depend on the outcomes in question, prediction markets aggregate the beliefs of the crowd by offering well-aligned financial incentives, allowing one to interpret the market price as a consensus prediction.

A particular prediction market framework has gained popularity in recent years, wherein participants trade not directly with each other but with a centralized computational agent called an automated market maker. We will see that while this automated framework enjoys many appealing properties, it lacks a crucial flexibility possessed by more traditional markets: the ability to adapt the magnitude of the market incentives to trading activity and external information shocks. In particular, we may wish the "depth" of the market to increase as more participants arrive, but decrease when information is released such as the outcome of a primary election.

In this talk, I will briefly discuss the history and theory behind prediction markets, and outline the potential-based automated market making framework due to Abernethy, Chen, and Wortman Vaughan. I will then give a range of theoretical findings about the design of adaptive markets using convex geometry, from impossibility results to new mechanisms and techniques. As we will see, although theoretical, these results directly inform the implementation of such automated markets, not only in the pricing mechanism, but in the design of the securities themselves.

Bio: Rafael Frongillo is an Assistant Professor of Computer Science at CU Boulder. His research lies at the interface between theoretical machine learning and economics, primarily focusing on domains such as information elicitation and crowdsourcing which involve the exchange of information for money, and drawing techniques from convex analysis, game theory, optimization, and statistics. Before coming to Boulder, Rafael was a postdoc at the Center for Research on Computation and Society at Harvard University and at Microsoft Research New York, and in 2013 received his Ph.D. in Computer Science at UC Berkeley, advised by Christos Papadimitriou and supported by the NDSEG Fellowship.
Prof. Soumik Sarkar Department of Mechanical Engineering, Iowa State University Spatiotemporal Graphical Modeling for Complex Cyber-Physical Systems Thursday, March 17th, 2016 3:30pm-5:00pm Clark's Room (Dean's Office)

Abstract: Modern distributed cyber-physical systems (CPSs) typically encounter a large variety of physical faults and cyber anomalies and in many cases, they are vulnerable to catastrophic fault propagation scenarios due to strong connectivity among the sub-systems. Majority of the state-of-the art techniques for detecting such anomalies depend on the knowledge of fault/attack characteristics. However, it is quite infeasible to have a comprehensive knowledge of faults/anomalies in real-life large CPSs. This talk will discuss a new data- driven modeling framework for system-wide anomaly detection in an unsupervised manner. The framework is based on a spatiotemporal feature extraction scheme built on the concept of symbolic dynamics for discovering and representing causal interactions among the subsystems of a CPS. The extracted spatiotemporal features are then used to learn system-wide patterns via a Restricted Boltzmann Machine (RBM) which is an energy based probabilistic graphical model. The anomaly detection process developed here aims to detect low probability events by using the concept of free energy of RBM.

Bio: Dr. Soumik Sarkar received his B. Eng. Degree in Mechanical Engineering in 2006 from Jadavpur University, Kolkata, India. He received M.S. in Mechanical Engineering and M.A. in Mathematics in 2009 from Penn State University. Dr. Sarkar received his Ph.D. in Mechanical Engineering from Penn State in 2011. He joined the Department of Mechanical Engineering at Iowa State as an Assistant Professor in Fall 2014. Previously, he was with the Decision Support & Machine Intelligence group at the United Technologies Research Center for 3 years as a Senior Scientist. Dr. Sarkar’s research interests include Statistical Signal Processing, Machine Learning, Sensor Fusion, Large volume information visualization, Fault Diagnostics & Prognostics, Distributed Control and Complexity Analysis with applications to complex cyber-physical systems, robotics, thermo-fluid sciences and plant science. He co‐authored 65 peer- reviewed publications including 23 journal papers, 3 book chapters and one magazine article. Dr. Sarkar is currently serving as an Associate Editor of Frontiers in Robotics and AI: Sensor Fusion and Machine Perception journal.
Dr. William Whitacre Draper Lab Automated 3D Digital Reconstruction of Fiber Reinforced Polymer Composites Tuesday, Feb 23th, 2016 3:30pm-5:00pm Clark's Room (Dean's Office)

Abstract: This presentation will be broken into two parts. The first part will provide an introduction to Draper Laboratory, including a brief history and summary of a few current projects. The second part of the talk will take a deep dive into an ongoing NASA project to develop virtual models of composite materials using X-ray computed tomography data (CT). This project is a combined experimental and computational effort aimed at high-resolution 3D imaging, visualization, and numerical reconstruction of a fiber-reinforced polymer (FRP) composite microstructure at the constituent-level length scale. A unidirectional sample of graphite/epoxy composite is imaged at sub-micron resolution using a 3D X-ray computed tomography microscope. The numerical reconstruction is enabled by a novel segmentation algorithm, which is developed using concepts adopted from computer vision and multi-target tracking, to detect and estimate with high accuracy the position of individual fibers in a volume of the imaged composite. A novel approach, based on the work of Julier and LaViola, is taken to include a nonoverlapping constraint into the estimation of the final reconstruction. The algorithm solves for an optimal, in the maximum a posteriori sense, adjustment for the virtual fiber locations and radii such that none of the virtual fibers overlap. Results from the segmentation algorithm are compared qualitatively to the tomographic data and suggest high accuracy of the numerical reconstruction.

Bio: Dr. William Whitacre is a Senior Member of the Technical Staff in the Perception and Localization Group at Draper Laboratory. His research supports a wide variety of programs ranging from strategic navigation and ballistic missile defense to X-ray computed tomography reconstruction of composite materials. Before joining Draper Laboratory, Dr. Whitacre was an Engineering Systems Architect in the Advanced Concepts and Technologies Division of Northrop Grumman Electronic Systems. While at Northrop Grumman he led a multi-year internally funded research and development project to create new sensor optimization and resource management approaches for distributed intelligence, surveillance, and reconnaissance applications; developed state-of-the-art multi-sensor, multi-target tracking algorithms to fuse data from networked radar sensors, infrared sensors, and electronic warfare sensors. He earned a PhD from Cornell University with his research on cooperative geolocation using UAVs with gimballing camera sensors. During his Ph.D. research, Dr. Whitacre worked with Insitu Inc. to implement a square root sigma point information filter for cooperative vision based geolocation of a moving ground target using the ScanEagle UAV. William is a member of the American Institute of Aeronautics and Astronautics and serves on the Guidance Navigation and Control Technical Committee.
Prof. Toru Namerikawa Department of System Design Engineering, Keio University, Yokohama, Japan Distributed Real-Time Pricing in Multi-period Electricity Market Tuesday, Jan 26th, 2016 3:30pm-5:00pm Clark's Room (Dean's Office)

Abstract: This talk deals with a distributed optimal power supply-demand management method based on dynamic pricing in the deregulated electricity market. Since power consumers and generators determine their own power demand or supply selfishly in the deregulated electricity market trading, some distributed power management methods are required to maintain the power supply-demand balance in a power grid. For this problem, the proposed method integrates two different time periods deregulated electricity market, "Day-ahead market" and "Real-time market", and solves this management problem in a distributed manner using electricity prices through market trading. Specifically, the proposed method, first, derives the optimal locational electricity prices which maximize social welfare of the entire power network in the day-ahead market based on alternating decision makings of market players. Then, the proposed method compensates the power imbalance caused by some problems such as prediction errors via negawatt trading in the real- time market, in which power consumers reduce their demand, while they receive monetary incentives from the market operator. The proposed method shows the optimal incentive design method using the day-ahead prices to minimize the power adjustment cost in real-time market trading. Finally, numerical simulation results are shown to demonstrate the effectiveness of the proposed method.

Bio: Toru Namerikawa received the B.E., M.E and Ph. D of Engineering degrees in Electrical and Computer Engineering from Kanazawa University, Japan, in 1991, 1993 and 1997, respectively. He is currently a Professor at Department of System Design Engineering, Keio University, Yokohama, Japan. He held visiting positions at Swiss Federal Institute of Technology in Zurich in 1998, University of California, Santa Barbara in 2001, University of Stuttgart in 2008 and Lund University in 2010. His main research interests are robust control, distributed and cooperative control and their application to mechatronic systems and power network systems.

Fall 2015

Speaker Institute Title Date Time Venue
Prof. Kira Barton Mechanical Engineering, Univerisity of Michigan Advancements in Modeling, Sensing and Control for High-Resolution Additive Manufacturing Thursday, Dec 3rd, 2015 11am-12:00pm DLC Collaboratory Bechtel Room

Abstract: Additive manufacturing (AM) describes a class of processes that perform a layer-by-layer “bottom-up” fabrication approach as opposed to traditional top-down, subtractive fabrication such as milling and lathing. Printing-based AM, and in particular micro-scale AM (µ-AM), has received significant attention in recent years as an enabling technology capable of revolutionizing the way we manufacture electronics, biosensors, and optics in this country. Meso-scale AM is capable of fabricating integrated features beyond what conventional machining can perform at this length scale. However, µ-AM has yet to demonstrate the fabrication of complex 3D structures at the micro-scale that are not fabricable by traditional micromachining. Limiting this step change in manufacturing capabilities is the reliance of μ-AM systems on a process monitoring, regulation, and quality control paradigm that is performed post-process and in an ad hoc manner. In this talk, we discuss some recent developments in process modeling, sensing, and control that aim to break this open-loop paradigm by providing the controls theoretic and process modeling knowledge to develop a robust closed-loop system for measurement and compensatory control.

Bio: Kira Barton is an Assistant Professor in the Department of Mechanical Engineering at the University of Michigan. She received her B.Sc. in Mechanical Engineering from the University of Colorado at Boulder in 2001. She continued her education in mechanical engineering at the University of Illinois at Urbana-Champaign and completed her M.Sc. and Ph.D. degrees in 2006 and 2010, respectively. She held a postdoctoral research position at the University of Illinois from Fall 2010 until Fall 2011, at which point she joined the Mechanical Engineering Department at the University of Michigan at Ann Arbor. Kira conducts research in modeling, sensing, and control for applications in advanced manufacturing and robotics, with a specialization in Iterative Learning Control and micro-additive manufacturing. Kira is the recipient of an NSF CAREER Award in 2014, 2015 SME Outstanding Young Manufacturing Engineer Award, and the 2015 University of Illinois, Department of Mechanical Science and Engineering Outstanding Young Alumni Award.
Prof. Eric Frew Aerospace Engineering, CU-Boulder Energy-Aware Unmanned Aircraft Systems: Enabling Persistent Atmospheric Sampling Thursday, Nov 19th, 2015 3:30pm-5:00pm Clark's Room (Dean's Office)

Abstract: The energy-aware airborne dynamic, data-driven application system (EA-DDDAS) performs persistent sampling in complex atmospheric conditions by exploiting wind energy using the dynamic data-driven application system paradigm. The main challenge for future airborne sampling missions is operation with tight integration of physical and computational resources over wireless communication networks, in complex atmospheric conditions. The physical resources considered here include sensor platforms, particularly mobile Doppler radar and unmanned aircraft, the complex conditions in which they operate, and the region of interest. Autonomous operation requires distributed computational effort connected by layered wireless communication. Onboard decision-making and coordination algorithms can be enhanced by atmospheric models that assimilate input from physics-based models and wind fields derived from multiple sources. These models are generally too complex to be run onboard the aircraft, so they need to be executed in ground vehicles in the field, and connected over broadband or other wireless links back to the field. Finally, the wind field environment drives strong interaction between the computational and physical systems, both as a challenge to autonomous path planning algorithms and as a novel energy source that can be exploited to improve system range and endurance. This seminar will describe a collaborative effort to implementation a complete EA-DDDAS. Results will be presented from previous field deployments of unmanned aircraft to show the evolution of the EA-DDDAS concept, and from recent deployments validating the EA-DDDAS.

Bio: Dr. Eric W. Frew is an associate professor in the Department of Aerospace Engineering Sciences and Director of the Research and Engineering Center for Unmanned Vehicles (RECUV) at the University of Colorado Boulder (CU). He received his B.S. in mechanical engineering from Cornell University in 1995 and his M.S and Ph.D. in aeronautics and astronautics from Stanford University in 1996 and 2003, respectively. Dr. Frew has been designing and deploying unmanned aircraft systems for over ten years. His research efforts focus on autonomous flight of heterogeneous unmanned aircraft systems; distributed information-gathering by mobile robots; miniature self-deploying systems; and guidance and control of unmanned aircraft in complex atmospheric phenomena. Dr. Frew was co-leader of the team that performed the first-ever sampling of a severe supercell thunderstorm by an unmanned aircraft. He is currently the CU Site Director for the National Science Foundation Industry / University Cooperative Research Center (IUCRC) for Unmanned Aircraft Systems. He received the NSF Faculty Early Career Development (CAREER) Award in 2009 and was selected for the 2010 DARPA Computer Science Study Group.
Dr. Emrah Akyol Coordinated Science Lab, University of Illinois Communication in Strategic Environments: Crawford-Sobel Meet Shannon Tuesday, Nov 10th, 2015 10:00am-11:00am DLC Collaboratory

Abstract: Over thirty years ago, economists Vincent Crawford and Joel Sobel introduced the concepts of strategic information transmission (SIT) and cheap talk in their seminal Econometrica paper, as a way of understanding how information is strategically revealed (or not) by agents whose interests are only partially aligned. This theory has had tremendous success in explaining situations ranging from advertising to expert advice sharing, and many extensions of the original SIT model and the broader “principal- agent” class of problems have been extensively studied in the economics literature since. However, despite its name and even superficially obvious connection with information theory (IT), SIT has so far received very little attention from the IT community.
In this talk, I will present approaches to address such strategic communication problems from the lens of information and game theories. Specifically, I will focus on a strategic communication paradigm where the better-informed transmitter communicates with a receiver who makes the ultimate decision concerning both agents. While the economists have extensively studied the Nash equilibrium variant of this problem, the more relevant Stackelberg equilibrium enables the use of Shannon theoretic tools. I will present the fundamental limits of strategic compression and communication problems in the SIT context. Particularly, three problem settings will be considered, focusing on the quadratic distortion measures and jointly Gaussian variables: compression, communication, and the simple equilibrium conditions without any compression or communication. The analysis will then be extended to the receiver side information setting, where the strategic aspect of the problem yields rather surprising results regarding optimality of single-letter, linear strategies. Finally, several applications of the results within the broader context of decision theory will be presented.
Joint work with Cedric Langbort and Tamer Basar of UIUC.

Bio: Emrah Akyol received the Ph.D. degree in 2011 from the University of California at Santa Barbara. From 2006 to 2007, he held positions at Hewlett-Packard Laboratories and NTT Docomo Laboratories, both in Palo Alto, CA where he worked on topics in video compression and streaming. From 2013 to 2014, Dr. Akyol was a postdoctoral researcher in the Electrical Engineering Department, University of Southern California. Currently, Dr. Akyol is a postdoctoral research associate in the Coordinated Science Laboratory, University of Illinois at Urbana-Champaign. His current research is on the interplay of networked information theory, game theory, communications, sensing and control. Dr. Akyol received the 2010 UCSB Dissertation Fellowship, the 2014 USC Postdoctoral Training Award and was an invited participant of the 2015 NSF Early-Career Investigators Workshop on CPS and Smart City.
Philip Brown ECEE, CU-Boulder Influencing Social Behavior: A Robust Approach Tuesday, Nov 3rd, 2015 3:30pm-5:00pm Clark's Room (Dean's Office)

Abstract: Uninfluenced social systems often exhibit suboptimal performance; accordingly, an important area of research deals with influencing mechanisms which aim to influence user behavior (by means of information, financial incentives, or otherwise), thereby bringing aggregate social behavior closer to optimal. In general, the efficiency guaranteed by a particular influencing mechanism is limited both by the quality of information available to the system-planner and the sophistication of the available influencing methodologies. If the planner possesses a perfect characterization of the system, it is often straightforward to perfectly align agents' incentives with the planner's global objective. However, as the quality of the planner's information decreases, increasingly sophisticated methodologies are required to achieve the same efficiency target. In this paper, we investigate situations in which the system-planner lacks such a perfect characterization and must employ methodologies that are robust to a variety of model imperfections.

Specifically, we study the application of taxes to a network-routing problem, and we assume that the tax-designer knows neither the network topology nor the tax-sensitivities and demands of the agents. We show that it is possible to design taxes that guarantee that network flows arising from selfish user behavior are arbitrarily close to optimal flows, despite the fact that users' tax-sensitivities are unknown to us. We term these taxes "universal," since they enforce optimal behavior in any routing game without requiring a priori knowledge of the specific game parameters. In general, these taxes may be very high; accordingly, for affine-cost parallel-network routing games, we explicitly derive the optimal bounded tolls and the best-possible efficiency guarantee as a function of a toll upper-bound. Finally, we show that if a system-planner has a limited ability to charge different taxes to different users on the basis of their tax-sensitivity, high-efficiency network flows can be enforced without employing large taxes.

Prof. Cesar O. Aguilar Math Department, Cal State Bakersfield Network Structure and Controllability in Multi-Agent Control Systems (ECEE Departmental Seminar) Tuesday, Oct 27th, 2015 11:00am-12:00pm DLC Collaboratory Bechtel Room (Engineering Center)

Abstract: Many engineering systems can be modeled as a collection of dynamic subsystems (or agents) that operate through local interactions via an information exchange network in order to accomplish system-level tasks. Examples of such systems include communication satellites, sensor networks for environmental monitoring, social interaction sites like Facebook and Twitter, biological processes such as yeast protein interactions and gene regulatory networks, parallel computing distributed systems, and many others. Although the presence of these so-called networked multi-agent systems in applications is diverse, many of them share the following common features: (1) the individual agents are dynamic systems with decision making capabilities, and (2) information can be diffused among the agents via an interconnection network. A central issue that binds these diverse multi-agent systems is to understand how the interconnection structure (local or global) of the underlying information network affects the dynamic properties and global behavior of the system. A fundamental property of a control system, both from an analytical and design perspective, is controllability, that is, the ability to transfer the system from one given state to another. Due to its importance, the controllability problem for multi-agent control systems has been an active research problem in the last decade. In this talk, I will present the state-of-the-art in this area and present new results that unifies several known conditions of controllability for multi-agent systems.

Bio: Cesar O. Aguilar received his undergraduate degree in Mathematics and Engineering in 2003 from Queen’s University (Canada), and his M.Sc. degree in Electrical and Computer Engineering in 2005 from the University of Alberta. He received the Ph.D. degree in Applied Mathematics from Queen’s University in 2010. He held a National Research Council postdoctoral fellowship in the Department of Applied Mathematics at the Naval Postgraduate School in Monterey, CA from 2010-2013. He is currently an Assistant Professor with the Department of Mathematics at California State University, Bakersfield. His research interests include nonlinear control, output regulation, optimal control, networked dynamical systems, and numerical methods in control theory.
Prof. Dan Szafir CS, CU-Boulder Human Interaction with Small Flying Robots Thursday, Oct 22nd, 2015 3:30pm-5:00pm Clark's Room (Dean's Office)

Abstract: Advances in microelectronics have enabled the design of a new class of robots with the ability to fly, ranging from consumer-grade quadcopters to NASA robots developed for the International Space Station. While such robots have captured public imagination, designing a robot that can fly does not necessarily produce a robot that is safe or useful for working with or near people. In this talk, I will discuss the promise of developing small flying robots that may act as collaborators in human environments and detail safety and usability challenges that stand in the way of this goal. I will then outline a design space for considering flying robots from a human-centered perspective and discuss several laboratory experiments that demonstrate the utility of this model in helping us overcome these challenges. In the course of this work, I have developed several empirically validated design prototypes that improve objective metrics of collaboration and user experiences when working with flying robots.

Bio: Daniel Szafir is an Assistant Professor within the ATLAS Institute and the Department of Computer Science at the University of Colorado Boulder. He received his Ph.D. in Computer Science from the University of Wisconsin–Madison. His work explores how we can leverage emerging interactive technologies, including small flying robots, wearable devices, and immersive virtual environments, to provide new forms of assistance to users in domains including collaborative work, education, and space exploration. His research support has included NASA, Google, and Mitsubishi Heavy Industries. His work been featured in several media outlets, including New Scientist, Engadget, and Discovery News. More information can be found on his website: http://danszafir.com
Dr. Ragavendran Gopalakrishnan XEROX Urban Mobility’s Green Future: Ridesharing and Electric Vehicles Thursday, Oct 15th, 2015 3:30pm-5:00pm Clark's Room (Dean's Office)

Abstract: This “works-in-progress” talk presents an overview of two projects underway at Xerox Research Centre India (XRCI) under the Sustainability theme: (1) First, we present a cost-sharing scheme for passengers who share (perhaps different portions of) a ride, either by carpooling, or using commercial cabs or ride-hailing services. Modeling a passenger’s disutilty as a combination of the monetary ride cost and the inconvenience cost due to detours, we provide constraints on the optimal route under which our cost-sharing scheme is Strongly/Sequentially Individually Rational (SIR), in the sense that not only are the passengers better off having chosen ridesharing, their disutility is non-increasing as the ride progresses and additional passengers are picked up. These SIR constraints are desirable for the passengers, because, coupled with a mobile app that can present this decrease in a visually compelling manner, our cost-sharing scheme has the potential to increase the adoption of ridesharing among commuters. (2) Second, we tackle the “chicken-and-egg” problem facing Electric Vehicle (EV) adoption–potential EV users are hesitant to make the switch because there isn’t enough charging infrastructure out there, and charging station network operators hesitate to set up charging stations if there is not sufficient demand for charging. Specifically, a charging station service provider is concerned with the demand at the candidate locations and the budget for deployment, whereas the potential EV user is concerned with charging station reachability and short waiting times at the station. We propose a ``mixed-packing-and-covering'' optimization framework that captures these competing concerns, and extends to an alternate scenario where a government agency seeks to optimize its allocation of grants (e.g., to incentivize setting up charging stations in low-demand areas) to competing network operators. We identify a family of heuristics that seeks to iteratively find the optimal allocation of the available budget towards satisfying the “packing” (e.g., budget) constraints and the “covering” (e.g., reachability) constraints by alternatingly invoking knapsack and set-cover algorithms. A final, non-technical part of the talk will provide a brief overview of XRCI and discuss internship/collaboration/grant opportunities for students and faculty.

Bio: Raga joined Xerox Research Centre India (XRCI) as a Research Scientist in the Algorithms and Optimization group in April this year. Prior to that, he was a postdoc with Jason Marden at ECEE, CU-Boulder. He obtained his Masters and PhD from Caltech in 2013. His research interests are broadly in the area of applied algorithmic game theory and optimization. At XRCI, Raga works on projects that are not only business-motivated and closer to the real-world, but also pose technical challenges with opportunities for rigorous research and performance modelling.
Prof. Emanuel Todorov CS, University of Washington Synthesis and stabilization of contact-rich behaviors with optimal control Tuesday, Sep 22nd, 2015 3:30pm-5:00pm Clark's Room (Dean's Office)

Abstract: Animals and machines interact with their environment mainly through physical contact. Yet the discontinuous nature of contact dynamics complicates planning and control, especially when combined with uncertainty. We have recently made progress in terms of optimizing complex trajectories that involve many contact events. These events do not need to be specified in advance, but instead are discovered fully automatically. Key to our success is the development of new models of contact dynamics, which enable continuation methods that in turn help the optimizer avoid a combinatorial search over contact configurations. We can presently synthesize humanoid trajectories in tasks such as getting up from the floor, walking and running, turning, riding a unicycle, as well as a variety of dexterous hand manipulation tasks. When augmented with warm-starts in the context of model predictive control, our optimizers can run in real-time and be used as approximately-optimal feedback controllers. Some of these controllers have already been transferred to physical robots, via ensemble optimization methods that increase robustness to modeling errors. The resulting trajectory libraries are also used to train recurrent neural networks. After training the networks can control the body autonomously, without further help from the trajectory optimizer.

Prof. Sean Humbert Mechanical Engineering, CU-Boulder Sensorimotor Integration and Control in Small-Scale Organisms Thursday, Sep 17nd, 2015 3:30pm-5:00pm Clark's Room (Dean's Office)

Abstract: One of the fundamental problems of interest to my laboratory is how to endow autonomous systems with the ability to make rapid control decisions in uncertain environments given limited computational resources and noisy sensory information. An example of significance is the operation of Unmanned Aircraft Systems (UAS) in urban settings, in particular the autonomous package delivery services proposed by Amazon PrimeAir and Google X Project Wing. As a source of inspiration, nervous systems have evolved to make useful reductions of sensory-rich and high dimensional data, forming simple representations along with feedback control paradigms that allow organisms to perform well with limited computation in the presence of uncertainty. My laboratory has applied control- and information-theoretic tools to formalize these principles in small organisms, providing insight into the biology and resulting in robust and computationally efficient solutions for small-scale engineered systems. In this talk we explore how sensorimotor principles from nature’s smallest fliers (insects) have been translated into novel solutions for perception and gust rejection for UAS, as well as other applications spanning navigation of Autonomous Underwater Vehicles (AUVs) to enhanced avionics systems for hypersonic vehicles.

Bio: J. Sean Humbert is the McLagan Family Endowed Associate Professor in the Department of Mechanical Engineering at CU Boulder. He holds a BS degree in Mechanical Engineering from the University of California, Davis, and MS and Ph.D. degrees in Mechanical Engineering from Caltech. Prof. Humbert’s research interests include bio-inspired robotics, estimation, and control theory, with applications to autonomous air, ground and underwater vehicles. Recent work focuses on perception and reduction principles of nervous systems, and agile flight control in insects and UAS. Dr. Humbert is a member of the Board on Army Science and Technology of the National Academies, is the recipient of the AIAA National Capital Section Hal Andrews Young Scientist/Engineer Award and an ARO Young Investigator Award. He is Director of the MAST-CTA Center on Microsystem Mechanics, Co-Director of the AFOSR Center of Excellence on Nature Inspired Flight Technologies, and a member of the Research and Engineering Center for Unmanned Vehicles (RECUV) at CU Boulder.
Dr. Oded Maler CNRS-VERIMAG, University of Grenoble, France Algorithmic Verification of Continuous and Hybrid Systems: Past, Present, Future Wednesday, Sep 9th, 2015 3:30pm-5:00pm Clark's Room (Dean's Office)
Dr. Oded Maler CNRS-VERIMAG, University of Grenoble, France Algorithmic Verification of Continuous and Hybrid Systems: Past, Present, Future Wednesday, Sep 9th, 2015 3:30pm-5:00pm Clark's Room (Dean's Office)

Abstract: Hybrid systems combine the qualitative discrete dynamics of transition systems with continuous dynamics as expressed by ordinary differential equations. They can be viewed either as continuous systems augmented with mode switching and if-then-else rules or as automata where in addition to the transition dynamics, there are real-valued variables that evolve while the automaton is idling inside a state. Analyzing the behavior of such systems is notoriously difficult. From the discrete verification standpoint we go out of the finite-state setting and have to deal with infinite-valued variables (and dense time). On the other hand, discontinuous mode switching often breaks down the analytical toolbox available for continuous systems, especially for linear systems.

In the first part of the talk I will talk about early work of the verification community on hybrid systems. This work can be characterized by:

The systems under study had a very simple (almost trivial) continuous dynamics in every state. Their simulation and analysis did not really require solving differential equations;

The question posed consisted of exact verification/reachability decision problems and a precise yes/no answers were sought.

I will present my own work on the topic that have shown that for a certain variant of these systems reachability can be decided in dimension 2 (two state variables) but not in higher dimensions.

In the second part of the talk I will illustrate contemporary algorithms which:

Can be applied to piecewise-affine dynamical systems, that is, hybrid automata such that the dynamics in each mode defined by a real linear differential equation;

Compute an over-approximation of the set of reachable states whose over-approximation error can be controlled.

These algorithms are implemented in the SpaceEx tool and can verify quite large systems. I will conclude with ongoing work on handling nonlinear systems and on moving closer to the only method currently used in practice, namely simulation. I will speculate about what is still missing in order to make this technology usable in practice.

Bio: Dr. Oded Maler is a research director at the CNRS (French national center for scientific research) working at the Verimag laboratory at Grenoble where he leads the hybrid systems group that made significant contributions to monitoring, systematic simulation and verification of systems the combine continuous dynamics (differential equations) and discrete events.
Dr. David Schlipf Stuttgart Lidar-assisted control for onshore and floating wind turbines Tuesday, Aug 25th, 2015 3:30pm-5:00pm Clark's Room (Dean's Office)

Abstract: Current advances in lidar technology provide opportunities to take a fresh look at wind turbine control. The wind is not only the main energy source but also the major disturbance to the control system. Thus, knowledge of the incoming wind is valuable information for optimizing energy production and reducing structural loads. Due to the measurement principles and the complexity of the wind, the disturbance cannot be measured perfectly. This forms a challenging task for estimation and control. The presentation will start with tailored wind and wind turbine models for lidar-assisted control concepts. Then, the estimation part will be addressed by model-based wind field reconstruction methods and by a spectral correlation model to optimize lidar scan configurations. In the control specific part, the presentation will discuss both a nonlinear feedforward controller applicable for onshore and floating wind turbines as well as a flatness-based feedforward controller for onshore wind turbines. Furthermore, initial results of the current field tests being conducted at the National Renewable Energy Laboratory (NREL) will be presented.

Bio: David Schlipf defended his PhD entitled “Lidar-Assisted Control Concepts for Wind Turbines” at the Stuttgart Wind Energy Institute, University of Stuttgart, where he is leading the research group “Control, Optimization and Monitoring”. Since October 2014, he has been a research scholar at the University of Colorado Boulder and NREL, carrying out field tests of lidar-assisted controllers and performing research in the floating wind turbine control area.
Prof. Naira Hovakimyan Mechanical Engineering, University of Illinois L1 Adaptive Control and Its Transition to Practice Thursday, Aug 20th, 2015 10:30pm-11:30pm Clark's Room (Dean's Office)

Abstract: The history of adaptive control systems dates back to early 50-s, when the aeronautical community was struggling to advance aircraft speeds to higher Mach numbers. In November of 1967, X-15 launched on what was planned to be a routine research flight to evaluate a boost guidance system, but it went into a spin and eventually broke up at 65,000 feet, killing the pilot Michael Adams. It was later found that the onboard adaptive control system was to be blamed for this incident. Exactly thirty years later, fueled by advances in the theory of nonlinear control, Air Force successfully flight tested the unmanned unstable tailless X-36 aircraft with an onboard adaptive flight control system. This was a landmark achievement that dispelled some of the misgivings that had arisen from the X-15 crash in 1967. Since then, numerous flight tests of Joint Direct Attack Munitions (JDAM) weapon retrofitted with adaptive element have met with great success and have proven the benefits of the adaptation in the presence of component failures and aerodynamic uncertainties. However, the major challenge related to stability/robustness assessment of adaptive systems is still being resolved based on testing the closed-loop system for all possible variations of uncertainties in Monte Carlo simulations, the cost of which increases with the growing complexity of the systems. This talk will give an overview of the limitations inherent to the conventional adaptive controllers and will introduce the audience to the L1 adaptive control theory, the architectures of which have guaranteed robustness in the presence of fast adaptation. Various applications, including flight tests of a Learjet, will be discussed during the presentation to demonstrate the tools and the concepts. With its key feature of decoupling adaptation from robustness L1 adaptive control theory has facilitated new developments in the areas of event-driven adaptation and networked control systems. A brief overview of initial results and potential directions will conclude the presentation.

Bio: Naira Hovakimyan received her MS degree in Theoretical Mechanics and Applied Mathematics in 1988 from Yerevan State University in Armenia. She got her Ph.D. in Physics and Mathematics in 1992, in Moscow, from the Institute of Applied Mathematics of Russian Academy of Sciences, majoring in optimal control and differential games. In 1997 she has been awarded a governmental postdoctoral scholarship to work in INRIA, France. In 1998 she was invited to the School of Aerospace Engineering of Georgia Tech, where she worked as a research faculty member until 2003. In 2003 she joined the Department of Aerospace and Ocean Engineering of Virginia Tech, and in 2008 she moved to University of Illinois at Urbana-Champaign, where she is a professor, university scholar and Schaller faculty scholar of Mechanical Science and Engineering. In 2015 she was named as inaugural director for Intelligent Robotics Lab of CSL at UIUC. She has co-authored a book and more than 300 refereed publications. She is the recipient of the SICE International scholarship for the best paper of a young investigator in the VII ISDG Symposium (Japan, 1996), the 2011 recipient of AIAA Mechanics and Control of Flight award and the 2015 recipient of SWE Achievement Award. In 2014 she was awarded the Humboldt prize for her lifetime achievements and was recognized as Hans Fischer senior fellow of Technical University of Munich. In 2015 she was recognized by UIUC Engineering Council award for Excellence in Advising. She is an associate fellow and life member of AIAA, a Senior Member of IEEE, and a member of SIAM, AMS, SWE, ASME and ISDG. Naira is co-founder of IntelinAir, Inc., a company that commercializes data-drones for various industries. Her research interests are in the theory of robust adaptive control and estimation, control in the presence of limited information, networks of autonomous systems, game theory and applications of those in safety-critical systems of aerospace, mechanical, electrical, petroleum and biomedical engineering.

Summer 2015

Speaker Institute Title Date Time Venue
Dr. Radhakishan Baheti National Science Foundation (NSF) NSF funding opportunities in Energy and Smart Systems Friday July 31st, 2015 11am-12:00pm Clark's Room (Dean's Office)

Abstract: The goal of the presentation is to provide an update on National Science Foundation (NSF) funding opportunities in Energy, Power, Control and Networks Program. The program deals with research and education in dynamical systems, networks, and control, smart grid, wind and solar energy, power electronics, and related areas. In addition, the cross disciplinary research in Cyber-Physical systems (CPS), and National Robotics Initiative (NRI) will be discussed. The focus of the CPS program is to develop the core system science needed to engineer complex cyber-physical systems upon which people can depend with high confidence. The CPS program brings together researchers from computations, communications, and control disciplines to address important engineering problems.

Bio: Dr. Kishan Baheti is a Program Director for Energy, Power, Control and Networks Program in the Division of Electrical, Communications, and Cyber Systems at the National Science Foundation.

Dr. Baheti received the B.S. and M.S. in Electrical Engineering in India from VRCE Nagpur, and from BITS Pilani, respectively. In 1970, he came to USA and received M.S. in Information and Computer Science from University of Oklahoma and Ph.D. in Electrical and Computer Engineering from Oregon State University. In 1976, Dr. Baheti joined the Control Engineering Laboratory of GE Corporate Research and Development Center in Schenectady, NY. His work focused on advanced multivariable control for jet engines, computer- aided control system design, vision-based robots for precision welding, and Kalman filtering. Dr. Baheti and his colleagues received IR-100 award for robotic welding vision system. He has organized a series of educational workshops for GE engineers that resulted in innovative product developments and contributed to enhance university collaborations with GE business divisions. In 1989, Dr.Baheti joined NSF as a Program Director in the Division of Electrical and Communications Systems. His contributions include the development of NSF initiatives on "Combined Research and Curriculum Development", "Semiconductor Manufacturing", and NSF/EPRI Program on "Intelligent Control". In addition, he started NSF Program "Research Experience for Teachers (RET)" to involve middle and high school teachers in engineering research that can be transferred to pre-college classrooms. Recently he is involved in cyber-physical systems, science of learning, and Robotics. He has served as associate editor for IEEE Transactions on Automatic Control, member of the Control Systems Board of Governors, chair for Public Information Committee, and awards chair for the American Automatic Control Council (AACC). He received "Distinguished Member Award" from the IEEE Control Systems Society. In 2013, he received “Outstanding Leadership and Service Award” from the Electrical and Computer Engineering Department Head Association. He was elected a Fellow of IEEE and a Fellow of AAAS.

Dr. Kendra Lesser Oxford Controller Synthesis and Safety Verification for Stochastic Systems Wednesday June 24th, 2015 3:30pm-5:00pm Clark's Room (Dean's Office)

Abstract: The focus of this talk is on safety verification and controller synthesis for stochastic systems, which can be addressed in the framework of stochastic reachability analysis. There are, however, some serious limitations in the applicability of reachability analysis to higher dimensional systems and systems that are partially observable. To address these issues, I will first discuss some alternative methods for computing stochastic reachable sets that are better suited to higher dimensional systems. I will then show how stochastic reachable sets can be used to solve motion planning problems in the presence of hundreds of stochastically moving obstacles. The rest of the talk will be concerned with partially observable systems, in which controllers must be constructed to satisfy safety specifications using only a stochastic observation process. I will show how abstraction techniques from the verification community can be combined with approximation algorithms for optimal control of partially observable Markov decision processes (POMDPs) to generate control laws that guarantee safety specifications are satisfied with maximum probability.

Bio: Kendra Lesser is currently a Marie Cure Fellow in the Department of Computer Science at the University of Oxford. She completed her PhD at the University of New Mexico in 2014, in the Department of Electrical and Computer Engineering. Her main research interests include formal verification and controller synthesis for partially observable systems, with the goal of making such analysis computationally feasible. Current applications include automated spacecraft rendezvous procedures and optimised maintenance and control of HVAC systems.
Prof. Justing Ruths Engineering Systems and Design, Singapore University of Technology and Design Control of Complex Networks Friday May 22nd, 2015 1:30pm-3:00pm Clark's Room (Dean's Office)

Abstract: Recent work at the borders of network science and control theory has begun to investigate the control of complex systems by studying their underlying network representations. In this talk I will present a new network statistic based on controllability, the control profile, that clarifies the underlying functional reasons for the placement of controls in a network. This statistic produces a mechanism to cluster networks into classes that are consistent with their large scale architecture and purpose. In the second half of this talk, I will present a new perspective on the control of networks motivated by modulating agent interactions rather than directly controlling agents, which leads to a bilinear control model. We develop graph-theoretic conditions for the structural control of a class of bilinear systems, which provide insight into the controllability conditions for classical, nonstructured systems.

Bio: Justin Ruths is an assistant professor and founding member of the Singapore University of Technology and Design, Singapore's new fourth national university, a collaborative venture with MIT. Justin holds degrees in Physics (BS, Rice University), Mechanical Engineering (MS, Columbia University), Electrical Engineering (MS, Washington University in Saint Louis), and Systems Science and Applied Math (PhD, Washington University in Saint Louis). His research themes include casting problems in the natural sciences and medicine as optimal control problems and investigating the control of large-scale systems. Towards this latter goal, some of his recent work is at the interface of control and network science.
Dr. Krishnamurthy Dvijotham Electrical Engineering, Caltech Differential Analysis the Power Flow Equations and Applications Thursday May 21st, 2015 3:30pm-5:00pm Clark's Room (Dean's Office)

Abstract: The AC power flow equations constitute a set of nonlinear equations. Finding solutions or determining the existence of a solution to these equations is hard in general, and studies have shown that they can exhibit complex and chaotic behavior. In this work, we apply the tools from monotone operator theory and robust control to the Jacobian of the power flow equations, and derive sufficient conditions for the existence and uniqueness of solutions to the power flow equation, and efficient algorithms to solve them. We discuss applications to loadability analysis, optimal power flow and state estimation in power systems.

Bio: Krishnamurthy Dvijotham is a postdoctoral fellow at the Center for Mathematics of Information at the California Institute of Technology. He received his PhD in Computer Science and Engineering from the University of Washington, Seattle in 2014 and his B.Tech from IIT Bombay in 2008. His current research focus is on optimization and control of large scale dynamical systems like the electric power grid and natural gas grid. He has also worked on machine learning and compressive sensing problems. He has won best student papers at the conference on Uncertainty in Artificial Intelligence (UAI 2014) and European Conference on Machine Learning (ECML 2008).

Spring 2015

Speaker Institute Title Date Time Venue
Prof. Karon MacLean Computer Science, University of British Columbia Physical Communication with Robots Tuesday Jan. 13th, 2015 3:30pm-5:00pm ECCR 265

Abstract: Buzzing cell phones and jolty game controllers are what the vast majority of users today think when they hear the word "haptics" (interaction through the sense of touch). One place where touch design has a very different role to play is in human-robot interaction. Here, advances in tactile sensing, wearable and context-aware computing and robotics more broadly are spurring new ideas about how to configure the human-robot relationship in terms of roles and utility, which in turn expose new technical and social design questions.

This talk will focus on my group’s recent work on haptic or physical human-robot interaction, where we aim to bring effective haptic interaction into people's lives by examining how touch (in either direction) can help address human needs with the benefit of both low- and high-tech innovation. I will give a sense of these efforts in two projects:

- Transparent physical communication with assistive robots, to support close collaboration in automation or home environments

- Affective physical communication with social robots - in both directions

Bio: Karon MacLean is Professor of Computer Science at the University of British Columbia, Canada, with a B.Sc. in Biology and Mech. Eng. (Stanford) and a M.Sc. and Ph.D. (Mech. Eng., MIT) and time spent as professional robotics engineer (Center for Engineering Design, University of Utah) and interaction researcher (Interval Research, Palo Alto). At UBC since 2000, her research specializes in haptic interaction: cognitive, sensory and affective design for people interacting with the computation we touch, emote and move with, whether robots, touchscreens or mobile activity sensors. She has innovated in human computer interaction curriculum design and teaching practices. Throughout her career MacLean has aspired to bridge the HCI, robotics and haptics communities, for example in helping to create the IEEE Transactions on Haptics (2008), reinventing the IEEE Haptics Symposium during 2010-2012, and offering many courses and journal special issues targeted at broad audiences. Peter Wall Early Career Scholar (2001); Izzak Walton Killam Memorial Faculty Research Fellowship (2007); Charles A. McDowell Award, 2008, NSERC Accelerator Award (2013) and numerous haptics, HCI and HRI editorial and advisory boards.
Prof. Ashutosh Nayyar Electrical Engineering, University of Southern California Sufficient statistics in Decentralized Decision-Making Problems Wedneseday Feb 18th, 2015 3:30pm-5:00pm Clark Room

Abstract: Decentralized systems are characterized by the presence of multiple decision-making agents acting on different information. Examples include sensor networks, teams of autonomous vehicles, power systems, networked control systems and communication networks. A key question in decentralized decision making problems is the following: Can the ever-growing information history available to agents/controllers be aggregated without compromising performance? In other words, are there sufficient statistics for the controllers? In this talk, we answer these questions for a general model of decentralized stochastic control problems where controllers sequentially share part of their information with one another. This model subsumes a large class of decentralized decision-making problems. We present a solution methodology for this model that is based on the concept of common information among the controllers. This common information approach provides a unified framework for several decentralized decision-making problems arising in diverse application domains.​

Bio: Ashutosh Nayyar received the B.Tech. degree in Electrical Engineering from the Indian Institute of Technology, Delhi, India. He received the MS and PhD degree in Electrical Engineering and Computer Science from the University of Michigan, Ann Arbor. He worked as a post-doctoral researcher at the University of Illinois at Urbana-Champaign and at the University of California, Berkeley before joining the University of Southern California as an assistant professor in 2014. His research focuses on the theory and applications of decentralized decision-making in a wide array of decentralized systems such as: sensing and communication systems, decentralized control systems, cyber-physical systems and electric energy systems.​
Prof. Girish Chowdhary Mechanical and Aerospace Engineering, Oklahoma State University Data-Driven Autonomous Monitoring of Massive Scaled Spatiotemporal Processes Tuesday Feb 24th, 2015 3:30pm-5:00pm Onizuka Room
Abstract: Unmanned Aerial Systems are envisaged to revolutionize monitoring of massive scale stochastic processes, such as the flux of CO_2 over landscapes above CO_2 storage sites, and dynamically evolving battlespaces. However, to realize this vision, UAS must autonomously adapt to real-world uncertainties and dynamical changes using streaming data in the presence of cyber-physical constraints. In this talk I will present an overview of new data-driven modeling and distributed monitoring paradigms for adaptive autonomy in uncertain spatiotemporally varying environments. On the data-driven modeling front, I will present a nonparametric modeling paradigm termed as Evolving - GP (E-GP), being designed for learning both abrupt and long-term changes in spatiotemporally evolving systems. On the distributed inference front, I will also present inference and UAS based sampling algorithms for distributed teams of mobile and static agents in the presence of cyber-physical constraints, such as limited communication range or flight-endurance. The new models and algorithms have been validated on real-world large datasets, and are expected to lead to autonomous UAS for modeling and monitoring massive scaled spatiotemporally evolving phenomena.
Bio: Girish Chowdhary is an assistant professor at Oklahoma State University, and the director of the Distributed Autonomous Systems laboratory at OSU. He holds a PhD from Georgia Institute of Technology. He was a postdoc at the Laboratory for Information and Decision Systems (LIDS) of the Massachusetts Institute of Technology for about two years. Prior to joining Georiga Tech, he also worked with the German Aerospace Center's (DLR's) Institute of Flight Systems for around three years. Girish's ongoing research interest is in theoretical insights and practical algorithms for adaptive autonomy over massive spatiotemporal scales, with a particular focus on applications in Unmanned Aerial Systems. He has authored over 70 peer reviewed publications in adaptive and fault tolerant control, sequential decision making and mission planning, aircraft system identification, distributed sensing and inference, Bayesian nonparametric learning for control, and vision aided navigation and control. He has led the development and flight-testing of over 10 research UAV platform. UAV autopilots based on Girish’s work have been designed and flight-tested on six UAVs, including by independent international institutions.
Prof. Sean Andersson Mechanical Engineering, Boston University Acquisition and analysis of single particle tracking data in fluorescence microscopy Tuesday March 3rd, 2015 3:30pm-5:00pm Onizuka Room
Abstract: Single Particle Tracking (SPT) is a powerful technique for understanding the dynamics in bimolecular systems, particularly inside living cells. By following the motion of single biological macromolecules (typically labeled with a fluorescence reporter), both qualitative (such as how the particle moves and what regions in the cell it explores) and quantitative (such as the values of diffusion coefficients and dwell times in particular locations) can be investigated. In this talk I will discuss two aspects of our work in the area of SPT. In the first part, I will present our confocal-based tracking scheme. While using a confocal (or multi-photon) imaging paradigm limits the field of view to essentially a single point, it provides benefits over wide field that include a better signal-to-noise ratio, typically better sensitivity, and a wider range along the optical axis. Our tracking algorithm, inspired by extremum seeking control, moves the detection volume of the microscope so as to move towards the peak of signal intensity. The technique will be described both mathematically and through experimental results. In the second part, I will turn to the analysis of SPT data. In the standard approach, individual particles are first localized in each image frame of a sequence and linked into a trajectory. That trajectory is then analyzed for motion characteristics. In general, however, the problems of localization and motion parameter estimation are coupled. I will describe our scheme for joint estimation of position and motion parameters. Through the use of Expectation-Maximization and Sequential Monte Carlo filtering, our scheme allows for general motion and observation models, allowing us to use complex, nonlinear descriptions that capture the behavior of the instrumentation and the particles being tracked.
Bio: Sean B. Andersson received a B.S. in engineering and applied physics (Cornell University, 1994), an M.S. in mechanical engineering (Stanford University, 1995), and a Ph.D. in electrical and computer engineering (University of Maryland, College Park, 2003). He has worked at AlliedSignal Aerospace and Aerovironment, Inc. and is currently an Associate Professor of mechanical engineering and of systems engineering with Boston University. His research interests include systems and control theory with applications in scanning probe microscopy, dynamics in molecular systems, and robotics.
Prof. Vijay Gupta Electrical Engineering, University of Notre Dame An Information Theoretic Notion of Security in Cyber-Physical Systems Tuesday March 10th, 2015 3:30pm-5:00pm Onizuka Room
Abstract: Cyber-physical systems are the next generation of engineering systems, with applications spanning critical infrastructure control, automotive systems, energy conservation, environmental monitoring, and robotics. One problem that is of interest in such systems is that of cyber-physical security. Specifically, it has been demonstrated that an intruder can hijack the information flow to degrade the estimation and control performance in such systems. I will show how concepts and tools from communication and information theory can be used to define a notion of security and characterize the control performance in the presence of degradations induced by malicious users through the communication networks. In a short second part, I will consider some other allied problems where similar tools may be useful.
Bio: Vijay Gupta received his B. Tech degree from the Indian Institute of Technology Delhi, and his M.S. and Ph.D . degrees from the California Institute of Technology, all in Electrical Engineering. He has served as a research associate at the Institute for Systems Research at the University of Maryland, College Park and as a consultant to the United Technologies Research Center. Since 2008, he has been in the Department of Electrical Engineering at the University of Notre Dame, where he is now an Associate Professor. He won the NSF CAREER award in 2009 and the Donald P. Eckman Award from the American Automatic Control Council in 2013. His research and teaching interests lie in the general area of intersection of control, communication and computation.
Prof. Eric Johnson School of Aerospace Engineering, Georgia Institute of Technology Navigation, Guidance, and Control to Enable Wider Use of Small Unmanned Aircraft Tuesday March 31st, 2015 3:30pm-5:00pm Onizuka Room
Abstract: The Georgia Institute of Technology Unmanned Aerial Vehicle Research Facility team is known for its research in the areas of navigation, guidance, and control; including flight testing. This work has included the first air launching of a hovering aircraft, the first automatic transition of an airplane to/from tail-sitting hover, vision-only formation flight, vision-aided inertial navigation, automatic helicopter flight with simulated stuck swash plate actuator, automatic airplane flight with half of one wing missing, and cooperative operations with multiple aircraft. The navigation, guidance, and control challenges for Unmanned Aerial Systems are among the most significant barriers to wider military or non-military use. This includes sensing and avoiding other aircraft and obstacles as well as navigation without reliance on Global Positioning Satellite systems. This presentation will address recent progress in the areas UAS navigation, guidance, and control, with emphasis on fault tolerant control, vision-aided inertial navigation, and laser-aided inertial navigation. This will include progress in both theory and related flight test validation on unmanned research aircraft.
Bio: Eric N. Johnson is the Lockheed Martin Associate Professor of Avionics Integration, Daniel Guggenheim School of Aerospace Engineering, Georgia Institute of Technology. He received a B.S. degree from University of Washington, M.S. degrees from MIT and The George Washington University, and a Ph.D. from Georgia Tech, all in Aerospace Engineering. He also has five years of industry experience working at Lockheed Martin and Draper Laboratory. As Georgia Tech faculty since 2001, he has performed research in adaptive flight control, aided inertial navigation, and autonomous systems. He was the lead system integrator for rotorcraft experiments for the DARPA Software Enabled Control program, which included the first air-launch of a hovering aircraft, automatic flight of a helicopter with a simulated frozen actuators, and adaptive flight control. He was the principal investigator of the Active Vision Control Systems AFOSR Multi-disciplinary University Research Initiative (MURI), which culminated in vision-based air-to-air tracking and vision-aided inertial navigation experimental validation. His most recent work has included automatic low altitude high speed flight of helicopters, indoor and outdoor vision-aided inertial navigation, fault tolerant control, and methods for sensing and avoiding other aircraft.
Jacob Aho ECEE, CU-Boulder Providing active power control services with wind energy for grid reliability Tuesday April 14st, 2015 3:30pm-5:00pm Onizuka Room
Abstract: As wind energy becomes a larger portion of the world’s energy portfolio there has been an increased interest for wind turbines to control their active power output to provide ancillary services which support grid reliability. This presentation will focus on the ability of wind energy to provide two of these services, primary frequency response, which consists of actuating to provide a stabilizing response to large disturbances to the grid, and frequency regulation, also referred to as secondary frequency control or automatic generation control (AGC), which provides minute-to-minute regulation of grid frequency. This talk will explain the control methodologies of providing these services with wind turbines, and discuss the performance of these services as well as the resulting damaging loads that are induced on the turbines. The final section of this talk will then focus on utilizing the frequency response and frequency regulation capabilities of wind energy when scheduling the power generation and ancillary service commitments of all generation on the grid.
Bio: Jacob Aho is a doctoral candidate in the Electrical, Computer, and Energy Engineering Department of the University of Colorado Boulder under the guidance of Professor Lucy Pao. He earned his B.S. and M.S. degrees in Electrical Engineering at the University of New Hampshire. Jacob’s research focus is on control systems that enable wind energy to provide active power grid reliability services and analyzing the optimal participation in these services. Jacob was born and raised in New Hampshire and enjoys many outdoor activities, such as hiking, camping, and hunting. Jacob also enjoys many sports, riding his motorcycle, and photography.
Prof. Roland Malhame Electrical Engineering, Polytechnique Montreal Mean Field Game Methods in the Control of Energy Systems Monday April 20th, 2015 3:30pm-5:00pm ECCR 211

Mean field game methodologies appeared about a decade ago as the result of independent research efforts in Applied Mathematics (Finance, Lions‐Lasry France), and Control Engineering ( Power control of cellular telephones, Huang‐Caines‐ Malhamé). A closely related concept, oblivious equilibrium, was also proposed in the context of Management Science (Weintraub, Benkard, Van Roy). Mean Field Games are essentially dynamic games involving a very large number of players, interacting anonymously, with the influence of each vanishing as their total number grows without bound. It is the major simplifications associated with the infinite population situation, used as a proxy to the large but finite population situation, which make this methodology so promising as far as applications are concerned. The potential applications range from reproducing with simple decentralized control laws herd and swarms dynamics, understanding opinions propagation, developing decentralized collective decision making mechanisms, understanding price fluctuations in an economy, etc.

In this talk, after a brief exposition of the main ideas behind Mean Field Game theory, we consider applications to a novel set of problems arising in energy systems. Due to increasing rates of penetration of intermittent renewable energy sources such as solar and wind energy in power systems, the fluctuations of instantaneous mismatches between generation and electricity demand have drastically augmented. Besides the attending network stability problems, this has deferred an increasing compensatory role to the so‐called spinning reserves in power systems; the latter typically rely on environmentally damaging fossil fuels. As an alternative, we aim at creating a control architecture that would allow the harnessing of the energy storage capability associated with millions of electrical devices such as electric water heaters, air conditioners, electric space heaters in dwellings and commercial buildings into a gigantic but distributed battery to smooth the generation‐load mismatch fluctuations, while maintaining local customer comfort and safety constraints. We develop novel formulations of the current mean field game theory that could help in achieving that goal. Numerical results are reported.

This is joint work with Arman Kizilkale.


Biography of Roland Malhamé Roland Malhamé received the Bachelor’s, Master’s and Ph.D. degrees in Electrical Engineering from the American University of Beirut, the University of Houston, and the Georgia Institute of Technology in 1976, 1978 and 1983 respectively.

After single year stays at University of Quebec , and CAE Electronics Ltd (Montreal), he joined in 1985 École Polytechnique de Montréal, where he is Professor of Electrical Engineering. In 1994, 2004, and 2012 he was on sabbatical leave respectively with LSS CNRS (France), École Centrale de Paris, , and University of Rome Tor Vergata.

His interest in statistical mechanics inspired approaches to the analysis and control of large scale systems has led him to contributions in the area of aggregate electric load modeling, and to the early developments of the theory of mean field games. His current research interests are in stochastic control, and the analysis and optimization of complex networks, in particular manufacturing, communication and power system networks. From june 2005 to june 2011, he was director of Groupe d’Études et de Recherche en Analyse des Décisions. He is an Associate Editor of International Transactions on Operations Research.

Prof. Sonia Martinez Department of Mechanical and Aerospace Engineering, UCSD Optimal Deployment for Mobile Robots in Constrained Scenario (Joint AES/CDS Seminar) Friday April 24th, 2015 2pm-3:00pm KOBL 330
Abstract: The distributed control and coordination of vehicles endowed with inexpensive sensing, communication, and computation devices has attracted an intense research activity in the last years. Just like animals do, these groups of mobile robots are envisioned to deploy over certain regions, assume certain specified patterns, or jointly move in a synchronized manner. However, these coordinated tasks are to be achieved with little available communication between different robots, no information about the global environment state, and different dynamic and operational restrictions. In this talk, we present an overview of the problem of optimal deployment for multi-agent systems and solutions building on Lloyd's algorithm that have been studied to address some of these questions. Then, we will focus on some recent work that deals with non-holonomic and environmental constraints, and how the self/event-triggered principle using sparse communications/computations can be exploited in this context.
Bio: Sonia Martínez is a Professor with the department of Mechanical and Aerospace Engineering at the University of California, San Diego. Dr. Martinez received her Ph.D. degree in Engineering Mathematics from the Universidad Carlos III de Madrid, Spain, in May 2002. Following a year as a Visiting Assistant Professor of Applied Mathematics at the Technical University of Catalonia, Spain, she obtained a Postdoctoral Fulbright Fellowship and held appointments at the Coordinated Science Laboratory of the University of Illinois, Urbana-Champaign during 2004, and at the Center for Control, Dynamical systems and Computation (CCDC) of the University of California, Santa Barbara during 2005. In a broad sense, Dr. Martínez' main research interests include the control of networked systems, multi-agent systems, nonlinear control theory, and robotics. For her work on the control of under-actuated mechanical systems she received the Best Student Paper award at the 2002 IEEE Conference on Decision and Control. She was the recipient of a NSF CAREER Award in 2007. For the paper "Motion coordination with Distributed Information," co-authored with Jorge Cortés and Francesco Bullo, she received the 2008 Control Systems Magazine Outstanding Paper Award. She has served on the editorial boards of the European Journal of Control (2011-2013), and currently serves on the editorial board of the Journal of Geometric Mechanics and IEEE Transactions on Control of Networked Systems.
Dr. Ahmad Beirami Department of Electrical and Computer Engineering, Duke University Quantifying computational security against brute-force attack Tuesday April 28th, 2015 3:30pm-5:00pm Onizuka Room

Despite several proposals for alternatives, passwords remain the primary means of securing online accounts in the cloud. The mathematical framework for quantifying the computational security of passwords against brute-force attack by query is formed by guesswork, which is the number of queries required of an adversary to breach a system by guessing a secret string. In this talk, we define "inscrutability" as the exponential rate of increase in average guesswork with respect to the secret string length and provide a finite-length approximation on inscrutability for finite-memory string sources. We also show that hiding the statistics of a finite-memory string source does not increase its inscrutability.

This talk is based on joint work with Robert Calderbank, Ken Duffy, and Muriel Medard.

Bio: Ahmad Beirami received his B.Sc. in Electrical Engineering from Sharif University of Technology in 2007 and his M.Sc. and Ph.D. in Electrical and Computer Engineering from Georgia Institute of Technology in 2011 and 2014, respectively. He is currently a postdoctoral scholar jointly affiliated with the information initiative at Duke (iiD) and the Research Laboratory of Electronics (RLE) at MIT. Beirami's Ph.D. work received the Center for Signal and Information Processing Outstanding Research Award (2014), the 2013-2014 school of ECE Graduate Research Excellence Award and the 2015 Sigma Xi Best Ph.D. Thesis Award all from Georgia Institute of Technology.

Fall 2014

Speaker Institute Title Date Time Venue
Prof. Sriram Sankaranarayanan Computer Science, CU-Boulder Statistical Techniques for Controller Performance Tuning Thursday Aug. 28th, 2014 3:30pm-5:00pm ECAD 150 (Clark)
Abstract: Controllers are often designed with tunable design parameters such as gains and thresholds that are adjusted for optimal performance. However, finding design parameters with robust performance guarantees is challenging in the presence of external stochastic disturbances, and plant parameter variations.
In this talk, we present a simulation-based approach to tackle the problem of finding values of design parameters that ensure bounds on performance measures with high probability. Our approach combines quantile regression to model the spread of the performance measures under stochastic parameter variations, with statistical hypothesis testing techniques to guarantee performance bounds with high probability. As such, the approach relies on repeated simulations, and is applicable to a variety of situations involving nonlinear systems. We will present an evaluation on a few benchmark problems and examine limitations of the approach.
Joint work with Yan Zhang and Fabio Somenzi.
Bio: Sriram Sankaranarayanan is an assistant professor of Computer Science at the University of Colorado, Boulder. His research interests include automatic techniques for reasoning about the behavior of computer and cyber-physical systems. Sriram obtained a PhD in 2005 from Stanford University where he was advised by Zohar Manna and Henny Sipma. Subsequently he worked as a research staff member at NEC research labs in Princeton, NJ. He has been on the faculty at CU Boulder since 2009. Sriram has been the recipient of awards including the President's Gold Medal from IIT Kharagpur (2000), Siebel Scholarship (2005), the CAREER award from NSF (2009) and the Dean's award for outstanding junior faculty for the College of Engineering at CU Boulder (2012).
Prof. Nisar Ahmed Aerospace Engineering, CU-Boulder Hybrid and Hierarchical Bayesian Data Fusion for Cooperative Human-Robot Perception Thursday Sept. 11th, 2014 3:30pm-5:00pm ECAD 150 (Clark)
Abstract: Ideally, intelligent autonomous robots should be able to reason about the uncertain world we live in completely on their own. This is difficult to do in practice for many reasons, e.g. computational limitations, which means that human reasoning must enter in at some level. To avoid undesirable task performance levels or unsafe situations for human-robot teaming, extensive research has shown that humans supervisors/teammates must learn appreciate their own physical and mental limits, as well as the limits of their machine counterparts. To successfully close the “human + robot” loop, truly intelligent systems must be designed with humans in mind as they become more sophisticated: robots should not only acknowledge and support human interaction, but also know how to best exploit human intelligence and capabilities whenever possible.

In this talk, I will present my ongoing work to develop formal Bayesian modeling and inference techniques that allow complex estimation problems to be solved via fusion of "soft data" (provided by humans) and "hard data" (provided by robots). In the context of target search applications with human-robot teams, I will show how to design generalized Gaussian sum Kalman filters for state estimation that listen to "everyday" semantic human language observations as well as to robot sensor data, using variational inference techniques for hybrid (discrete and continuous) random variables. I will also show how poorly characterized "human sensor" data can be simultaneously calibrated and fused online via hierarchical Bayesian inference techniques. Time permitting, I will also describe how these data fusion techniques can be scaled up to large multi-human/multi-robot networks via decentralized Bayesian message-passing.
Bio: Nisar Ahmed is an assistant professor of Aerospace Engineering Science at the University of Colorado Boulder. His research interests are in modeling and estimation for intelligent control of dynamical systems, especially for applications involving human-robot interaction, sensor networks and information fusion. He completed his Ph.D. in Mechanical Engineering at Cornell University in 2012 and from 2012-2014 was a postdoctoral research associate in the Cornell Autonomous Systems Lab (ASL) with Professor Mark Campbell. He was awarded an NSF Graduate Research Fellowship in 2007, the 2011 AIAA GNC Conference Best Paper Award, and an ASEE Air Force Summer Faculty Fellowship in 2014.
Prof. Jason Marden Electrical Engineering, CU-Boulder The Role of Information in Multiagent Coordination Tuesday Sept. 23rd, 2014 3:30pm-5:00pm ECAD 150 (Clark)
Abstract: The goal in networked control of multiagent systems is to derive desirable collective behavior through the design of local control algorithms. The information available to the individual agents, either attained through communication or sensing, invariably defines the space of admissible control laws. Hence, informational restrictions impose constraints on achievable performance guarantees. The first part of this talk will provide one such constraint with regard to the efficiency of the resulting stable solutions for a class of networked resource allocation problems with submodular objective functions. When the agents have full information regarding the mission space, the efficiency of the resulting stable solutions is guaranteed to be within 50% of optimal. However, when the agents have only localized information about the mission space, which is a common feature of many well-studied control designs, the efficiency of the resulting stable solutions can be 1/n of optimal, where n is the number of agents. Consequently, in general such schemes cannot guarantee that systems comprised of n agents can perform any better than a system comprised of a single agent for identical environmental conditions. The second part of this talk will focus on identifying how augmenting the information to the agents can impact achievable performance guarantees in multiagent systems. Clearly, providing the agents with additional information can lead to control designs with improved efficiency guarantees associated with the emergent collective behavior. However, we demonstrate that such gains frequently come at the expense of the underlying convergence rate.
Bio: Jason Marden is an Assistant Professor in the Department of Electrical, Computer, and Energy Engineering at the University of Colorado. Jason received a BS in Mechanical Engineering in 2001 from UCLA, and a PhD in Mechanical Engineering in 2007, also from UCLA, under the supervision of Jeff S. Shamma, where he was awarded the Outstanding Graduating PhD Student in Mechanical Engineering. After graduating from UCLA, he served as a junior fellow in the Social and Information Sciences Laboratory at the California Institute of Technology until 2010 when he joined the University of Colorado. Jason is a recipient of the NSF Career Award (2014), the AFOSR Young Investigator Award (2012), and the American Automatic Control Council Donald P. Eckman Award (2012). Jason's research interests focus on game theoretic methods for the control of distributed multiagent systems.
Prof. William Sandholm Economics, University of Wisconsin Large Deviations and Stochastic Stability in Single and Double Limits Oct. 7th, 2014 3:30pm-5:00pm ECAD 150 (Clark)
Abstract: We consider a model of stochastic evolution under general noisy best response protocols, allowing the probabilities of suboptimal choices to depend on their payoff consequences. Our analysis focuses on behavior in the small noise double limit: we first take the noise level in agents' decisions to zero, and then take the population size to infinity. We show that in this double limit, escape from and transitions between equilibria can be described in terms of solutions to continuous optimal control problems. These are used in turn to characterize the asymptotics of the the stationary distribution, and so to determine the stochastically stable states. We then focus on a class of examples - three-strategy coordination games that satisfy the marginal bandwagon property and that have an interior equilibrium, and choices governed by the logit rule. In this setting the control problems can be solved explicitly, and the analysis carried to its very end. We argue that this tractability should persist under other classes of games and other choice rules.
Bio: William H. Sandholm is Professor of Economics at the University of Wisconsin-Madison. His research is in game theory, with a focus on deterministic and stochastic disequilibrium dynamics. He is the author of "Population Games and Evolutionary Dynamics" (MIT Press, 2010), and is a member of the Council of the Game Theory Society.
Prof. Edwin Chong Electrical Engineering, Colorado State University Decision Making in Large Networks Wed. Oct. 22nd, 2014 3:00pm-4:00pm ECCR 1B55
Abstract: Is there wisdom in crowds? What kind of crowds? How should individuals share their knowledge? What could go wrong? These questions are relevant to decision-making issues in social networks and sensor networks. We consider large networks with two specific topologies: hierarchical networks, common in enterprise and military settings, and feedforward networks, common in social networks with sequential information sharing. We consider a hypothesis testing problem and study the dependence of "global" (public) decisions on several factors: network topology, richness of "private" information, richness of shared information, and errors in shared information. Our results expose the crucial role of network topology on the reliability of the wisdom of crowds.
Bio: Edwin K. P. Chong received the M.A. and Ph.D. degrees in 1989 and 1991, respectively, both from Princeton University, where he held an IBM Fellowship. He then joined the School of Electrical and Computer Engineering at Purdue University in 1991. Since August 2001, he has been a Professor of Electrical and Computer Engineering and a Professor of Mathematics at Colorado State University. His current interests are in networks and optimization methods. He coauthored the best-selling book, An Introduction to Optimization (4th Ed., Wiley-Interscience, 2013). He is currently a Senior Editor of the IEEE Transactions on Automatic Control, and formerly an editor for Computer Networks and the Journal of Control Science and Engineering. He is a Fellow of the IEEE, and served as an IEEE Control Systems Society Distinguished Lecturer. He received the NSF CAREER Award in 1995 and the ASEE Frederick Emmons Terman Award in 1998. He was a co-recipient of the 2004 Best Paper Award for a paper in the journal Computer Networks. In 2010, he received the IEEE Control Systems Society Distinguished Member Award. He is currently the Vice President for Financial Activities for the Control Systems Society. He has served as Principal Investigator for numerous funded projects from NSF, DARPA, and other federal funding agencies, and from industrial sources.
Prof. Juan G. Restrepo Applied Math, CU-Boulder Critical dynamics in networks of excitable systems with inhibition Oct. 28th, 2014 3:30pm-5:00pm ECAD 150 (Clark)
Abstract: It has been hypothesized that the brain operates in a “critical" regime in which excitation and inhibition are precisely balanced. Experimental signatures of critical dynamics in functional brain networks have been successfully modeled using networks of simple excitable systems. However, these models do not typically include inhibitory nodes and cannot sustain critical dynamics without external stimulation. We introduce a model with inhibitory nodes and find that it can sustain critical dynamics with a lifetime that grows exponentially with (Na/k), where N is the number of nodes, a is the fraction of inhibitory nodes, and k is the mean degree of the network. For relevant values of N, a, and k, the critical dynamics are effectively ceaseless. Our analysis is based on the “branching function”, which represents the expected fractional growth in the number of active nodes per time step. The statistics of avalanche sizes observed in our model agrees with that obtained experimentally.
Bio: Juan G. Restrepo received the B.S. degree in physics from the University of Los Andes, Bogotá, Colombia, in 1999 and the Ph.D. degree in Applied Mathematics from the University of Maryland in 2005. He is currently an Assistant Professor with the Department of Applied Mathematics at CU Boulder. His research interests include dynamical systems, chaos, synchronization of coupled oscillators, dynamics on complex networks, and cardiac dynamics.
Prof. Gabe Sibley Computer Science, CU-Boulder Mobile Robot Perception for Long-term Autonomy Nov. 4th, 2014 3:30pm-5:00pm ECAD 150 (Clark)
Abstract: If mobile robots are to become ubiquitous, we must first solve fundamental problems in perception. Before a mobile robot system can act intelligently, it must be given -- or acquire -- a representation of the environment that is useful for planning and control. Perception comes before action, and the perception problem is one of the most difficult we face. An important goal in mobile robotics is the development of perception algorithms that allow for persistent, long-term autonomous operation in unknown situations (over weeks or more). In our effort to achieve long-term autonomy, we have had to solve problems of both metric and semantic estimation. In this talk I will describe two recent and interrelated advances in robot perception aimed at enabling long-term autonomy. The first is relative bundle adjustment (RBA). By using a purely relative formulation, RBA addresses the issue of scalability in estimating consistent world maps from vision sensors. In stark contrast to traditional SLAM, I will show that estimation in the relative framework is constant-time, and crucially, remains so even during loop-closure events. This is important because temporal and spatial scalability are obvious prerequisites for long-term autonomy. Building on RBA, I will then describe co-visibility based place recognition (CoVis). CoVis is a topo-metric representation of the world based on the RBA landmark co-visibility graph. I will show how this representation simplifies data association and improves the performance of appearance based place recognition. I will introduce the "dynamic bag-of-words" model, which is a novel form of query expansion based on finding cliques in the co-visibility graph. The proposed approach avoids the -- often arbitrary -- discretization of space from the robot's trajectory that is common to most image-based loop-closure algorithms. Instead, I will show that reasoning on sets of co-visible landmarks leads to a simple model that out-performs pose-based or view-based approaches, in terms of precision and recall. In summary, RBA and CoVis are effective representations and associated algorithms for metric and semantic perception, designed to meet the scalability requirements of long-term autonomous navigation.
Bio: Gabe Sibley is an assistant professor in Computer Science at University of Colorado, Boulder. He was formerly an assistant professor at GWU and a member of the University of Oxford Mobile Robotics Group. He did his PhD at the University of Southern California and at NASA-JPL, where he worked on long-range data-fusion algorithms for planetary landing vehicles, unmanned sea vehicles and unmanned ground vehicles. His core interest is in probabilistic perception algorithms and estimation theory that enable long-term autonomous operation of mobile robotic systems, particularly in unknown environments. He has extensive experience with vision based, real-time localization and mapping systems, and is interested in fundamental understanding of sufficient statistics that can be used to represent the state of the world. His research uses real-time, embodied robot systems equipped with a variety of sensors -- including lasers, cameras, inertial sensors, etc. -- to advance and validate algorithms and knowledge representations that are useful for enabling long-term autonomous operation.
Dr. Timothy Caldwell CS, CU-Boulder A Planner for Complex Dynamic Systems Nov. 11th, 2014 3:30pm-5:00pm ECAD 150 (Clark)
Abstract: The talk presents an algorithm for planning around obstacles for complex dynamical systems. In order to enforce the dynamics, the planner must be designed to handle the intrinsic numerical issues of unstable or chaotic systems. In the talk I extend the popular rapidly exploring random tree algorithm to complex dynamic systems so that the planner 1) efficiently explores the dynamics—i.e. distance for the nearest neighbor calculation is the optimal controls cost—2) addresses numerical issues associated with sensitivities to initial conditions, and 3) maintains low computational complexity. My algorithm approximates the reachable states through a linearization and the reachability Grammian and enforces the dynamics through a projection. A trust region-like approach maintains the quality of linearization approximation. Additionally, through pre-computations and caching, solving the optimal control problem for determining the nearest neighbor reduces to matrix manipulations as opposed to iteratively solving state and co-state differential equations. Examples are shown for planning the n-link inverted pendulum through a corridor.
Bio: Timothy M. Caldwell is a Postdoctoral Researcher in the Computer Science Department at the University of Colorado as a member of the Correll Lab. He obtained his PhD at Northwestern University in 2013 as a Department of Energy Office of Science Graduate Fellow.
Holly Borowski ECEE, CU-Boulder Learning Efficient Correlated Equilibria Nov. 18th, 2014 3:30pm-5:00pm ECAD 150 (Clark)
Abstract: The majority of distributed learning literature focuses on convergence to Nash equilibria. Correlated equilibria, on the other hand, can characterize more efficient collective behavior than even the best Nash equilibrium. However, there are no existing distributed learning algorithms that converge to specific correlated equilibria. This work provides one such algorithm which guarantees that the agents' collective joint strategy will constitute an efficient correlated equilibrium with high probability. The key to attaining efficient correlated behavior through distributed learning involves incorporating a common random signal into the learning environment.
Bio: Holly is a PhD student with Jason Marden. Her research focuses on the impact of information on distributed systems. In particular, she is investigating how information communicated between agents impacts convergence rates and efficiency in distributed control algorithms.
Prof. Francois Meyer ECEE, CU-Boulder Random Graph Models for Image Patches Dec. 2nd, 2014 3:30pm-5:00pm ECAD 150 (Clark)

In this talk we address the problem of understanding the success of algorithms that organize patches according to graph-based metrics. Algorithms that analyze patches extracted from images or time series have led to state-of-the art techniques for classification, denoising, and the study of nonlinear dynamics.

In the first part of the talk, we construct random graph models for image patches extracted from natural images. These graphs epitomize the geometry observed in general patch-graphs. We provide a detailed analysis of the commute time metric on such graphs. We then study the eigenvectors of the graph Laplacian, and prove that a parametrization of the graph based on commute times shrinks the mutual distances between patches that correspond to rapid local changes in the signal, while the distances between patches that correspond to slow local changes expand.

In effect, our results explain why the parametrization of the set of patches based on the eigenfunctions of the Laplacian can concentrate patches that correspond to rapid local changes, which would otherwise be scattered in the space of patches. While our results are based on a large sample analysis, numerical experimentations on synthetic and real data indicate that the results hold for datasets that are very small in practice.


Francois Meyer graduated with Honors from Ecole Nationale Superieure d’Informatique et de Mathematiques Appliquees, Grenoble, in 1987, with a M.S. in applied mathematics. He received a Ph.D. degree in electrical engineering from INRIA, France, in 1993. He is currently a Professor with the Department of Electrical Engineering, University of Colorado, Boulder. He had previously been an Assistant Professor at Yale University (1997-1999), a Visiting Professor at the Institute Henri Poincaré, Paris (1999), a Senior Fellow at the Institute of Pure and Applied Mathematics, (UCLA) (2004), a Visiting Research Scholar at Princeton University (2007), and a Visiting Scholar at the Institute for Computational and Experimental Research in Mathematics, Brown University (2014).

Meyer and his students work on the development of mathematical algorithms and computational methods for the analysis of observational high-dimensional dataset; application to biology, neuroscience (fMRI, EEG), geoscience, etc.

Dr. Soomin Lee Mechanical Engineering, Duke University Distributed Optimization Over Decentralized Network Systems Dec. 9th, 2014 3:30pm-5:00pm ECAD 150 (Clark)
Abstract: We witness a growing interest in distributed multi-agent systems. The Internet, electric power systems, mobile communication networks, privacy aware networks and social networks are just a few examples of the myriad network systems that have become a part of everyday life for many people. Lots of interesting optimization problems arise in such network systems. The agents on these networks usually have distributed problem data, but in practice there is no data fusion center that can see the problem as a whole, gather the information globally, or synchronize actions. Furthermore, the network agents might have varying restrictions on energy, data storage and computational capabilities. In this talk, I will present efficient decentralized and distributed optimization algorithms for such systems that allow the network agents to achieve provable consensus to the global optimum. In particular, our primary concerns are to understand the system dynamics under local and decentralized operations, to implement localized communication protocols which can process the distributed information efficiently and robustly, and to develop low-memory, computationally light distributed optimization techniques. I will also present some applications of the algorithms in various engineering disciplines. The future vision and possible extension of this work will be discussed as well.
Bio: Dr. Soomin Lee is currently working as a Postdoctoral Associate in Mechanical Engineering and Materials Science at Duke University. She received her Ph.D. in Electrical and Computer Engineering from the University of Illinois, Urbana-Champaign (2013). She received two master's degrees from the Korea Advanced Institute of Science and Technology in Electrical Engineering, and from the University of Illinois at Urbana-Champaign in Computer Science. In 2009, she was an assistant research officer at the Advanced Digital Science Center (ADSC) in Singapore. Her research interests include theoretical optimization, control and optimization of various distributed engineering systems interconnected over complex networks, risk-averse modeling of multiagent robotic systems under dynamically changing and uncertain environments.


To subscribe/unsubscribe to our mailing list, please follow the instruction at: http://www.colorado.edu/oit/tutorial/email-list-manager-subscribe-list