Safe Learning and Control with L1 Adaptation

February 26, 2021, 2:00 – 4:00 pm

Naira Hovakimyan

W. Grafton and Lillian B. Wilkins Professor of Mechanical Science and Engineering at UIUC.


Learning-based control paradigms have seen many success stories with various robots and co-robots in recent years. However, as these robots prepare to enter the real world, operating safely in the presence of imperfect model knowledge and external disturbances is going to be vital to ensure mission success. In the first part of the talk, we present an overview of L1 adaptive control, how it enables safety in autonomous robots, and discuss some of its success stories in the aerospace industry. In the second part of the talk, we present some of our recent results that explore various architectures with L1 adaptive control while guaranteeing performance and robustness throughout the learning process. An overview of different projects at our lab that build upon this framework will be demonstrated to show different applications.


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 from the Institute of Applied Mathematics of Russian Academy of Sciences in Moscow. She is currently a W. Grafton and Lillian B. Wilkins Professor of Mechanical Science and Engineering at UIUC. In 2015 she was named inaugural director for Intelligent Robotics Lab of Coordinated Science Laboratory at UIUC. She has co-authored two books, eleven patents and more than 450 refereed publications. She was 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, the 2015 recipient of SWE Achievement Award, the 2017 recipient of IEEE CSS Award for Technical Excellence in Aerospace Controls, and the 2019recipient of AIAA Pendray Aerospace Literature Award. In 2014 she was awarded the Humboldt prize for her lifetime achievements. She is Fellow and life member of AIAA and a Fellow of IEEE. She is cofounder and chief scientist of IntelinAir. Her work in robotics for elderly care was featured in the New York Times, on Fox TV and CNBC. Her research interests are in control, estimation and optimization, autonomous systems, game theory and their broad applications across various industries.

Bluejeans link:

Enabling Human-Aware Automation: A Dynamical Systems Perspective on Human Cognition

November 13, 2020 – 11:00 AM

Neera Jain

Perdue University


Across many sectors, ranging from manufacturing to healthcare to the military theater, there is growing interest in the potential impact of automation that is truly collaborative with humans. Realizing this impact, though, rests on first addressing the fundamental challenge of designing automation to be aware of, and responsive to, the human with whom it is interacting. While a significant body of work exists in intent inference based on human motion, a human’s physical actions alone are not necessarily a predictor of their decision-making. Indeed, cognitive factors, such as trust and workload, play a substantial role in their decision making as it relates to interactions with autonomous systems. In this talk, I will describe our interdisciplinary efforts at tackling this problem, focusing on recent work in which we synthesized a near-optimal control policy using a trust-workload POMDP (partially-observable Markov decision process) model framework that captures changes in human trust and workload for a context involving interactions between a human and an intelligent decision-aid system. Using automation transparency as the feedback variable, we designed a policy to balance competing performance objectives in a reconnaissance mission study in which a virtual robotic assistant aids human subjects in surveying buildings for physical threats. I will present experimental validation of our control algorithm through human subject studies and highlight how our approach is able to mitigate the negative consequences of “over trust” which can occur in such interactions. I will also briefly discuss our related work involving the use of psychophysiological data and classification techniques as an alternative method toward real-time trust estimation.


Dr. Neera Jain is an Assistant Professor in the School of Mechanical Engineering and a faculty member in the Ray W. Herrick Laboratories at Purdue University. She directs the Jain Research Laboratory with the aim of advancing technologies that will have a lasting impact on society through a systems-based approach, grounded in dynamic modeling and control theory. A major thrust of her research is the design of human-aware automation through control-oriented modeling of human cognition. A second major research thrust is optimal design and control of complex energy systems. Dr. Jain earned her M.S. and Ph.D. degrees in mechanical engineering from the University of Illinois at Urbana-Champaign in 2009 and 2013, respectively. She earned her S.B. from the Massachusetts Institute of Technology in 2006. Upon completing her Ph.D., Dr. Jain was a visiting member of the research staff at Mitsubishi Electric Research Laboratories where she designed model predictive control algorithms for HVAC systems. In 2015 she was a visiting summer researcher at the Air Force Research Laboratory at Wright-Patterson Air Force Base. Dr. Jain and her research have been featured in NPR and Axios. As a contributor for, she writes on the topic of human interaction with automation and its importance in society. Her research has been supported by the National Science Foundation, Air Force Research Laboratory, Office of Naval Research, as well as private industry.

Three Problems in Mathematical Oncology

October 16, 2020 – 03:15 PM

Paul K. Newton

University of Southern California


I will introduce three problems in mathematical oncology all of which involve nonlinear dynamics and control theory. First, I will describe our work using Markov chain models to forecast metastatic progression. The models treat progression as a (weighted) random walk on a directed graph whose nodes are tumor locations, with transition probabilities obtained through historical autopsy date (untreated progression) and longitudinal data (treated) from Memorial Sloan Kettering and MD Anderson Cancer Centers. Then, I will describe our models that use evolutionary game theory (replicator dynamics with prisoner’s dilemma payoff matrix) to design multi-drug adaptive chemotherapy schedules to mitigate chemo-resistance by suppressing ‘competitive release’ of resistant cell populations. The models highlight the advantages of antagonistic drug interactions (over synergistic ones) in shaping the fitness landscape of co-evolving populations. Finally, I will describe our work on developing optimal control schedules (based on Pontryagin’s maximum principle) that maximize cooperation for prisoner’s dilemma replicator dynamical systems. As much as possible with the Zoom format, I hope the seminar will be interactive and a starting point for further discussions.


Professor Newton received his B.S. (cum laude) degree in Applied Mathematics/Physics at Harvard University in 1981 and his Ph.D. in 1986 from the Division of Applied Mathematics at Brown University. He then moved to the Mathematics Department at Stanford University to work as a post-doctoral scholar under J.B. Keller. He became Assistant (1987) and Associate Professor (1993) in the Mathematics Department at the University of Illinois Champaign-Urbana (UIUC) and at the Center for Complex Systems Research (CCSR) at the Beckman Institute. In 1993 he moved to the Aerospace & Mechanical Engineering Department and the Mathematics Department at the University of Southern California and was promoted to Full Professor in 1998. Trained as an applied mathematician, Professor Newton’s work focuses on developing mathematical models for nonlinear dynamical processes in continuum mechanics and biophysics, currently focusing mostly on mathematical oncology and systems biology. He has held visiting appointments at Caltech, Brown, Hokkaido University, The Kavli Institue for Theoretical Physics at UC Santa Barbara, and The Scripps Research Institute where he functioned as head of the mathematical modeling section of the NCI supported Physical Sciences Oncology Center (2009-2014). He is currently a Professor of Applied Mathematics, Engineering, and Medicine in the Viterbi School of Engineering, the Dornsife College of Letters, Arts and Sciences, the Norris Comprehensive Cancer Center in the Keck School of Medicine, and a founding affiliate member of the LJ Ellison Institute for Transformative Medicine of USC. He currently serves as Editor-in-Chief of the Journal of Nonlinear Science (SpringerNature).

Stochastic Approximation: Some New Wine in Old Bottle

October 06, 2020 – 11:00 AM

Vivek S. Borkar

Indian Institute of Technology Bombay


This talk will give an overview of old and new results and directions in stochastic approximation algorithms, broadly split into basic theory, variants, and applications. Central to all this will be the ‘o.d.e’ (for ‘Ordinary Differential Equations’) approach to their analysis.


Prof. Vivek S. Borkar is CSIR Bhatnagar Emeritus Fellow at Indian Institute of Technology Bombay. He obtained his B.Tech. (EE) from IIT Bombay, M.S. (Systems and Control) from Case Western Reserve Uni., and Ph.D. (EECS) from the Uni. of California, Berkeley, in 1976, 77, 80 resp. He has held positions in TIFR Centre and Indian Institute of Science, Bengaluru, and Tata Inst. of Fundamental Research and IIT Bombay in Mumbai. He is a Fellow of IEEE, AMS, TWAS and various science and engineering academies in India. He was awarded the S. S. Bhatnagar Prize in engineering sciences by the Government of India in 1992 and was an invited speaker at the International Congress of Mathematicians in Madrid in 2006. His research interests are in stochastic control and optimization, inclusive of theory, algorithms, and applications, particularly to communications.

Mean Field Differential Games with Elements of Robustness

September 04, 2020 – 02:00 PM

Tamer Basar

University of Illinois, Urbana


Perhaps the most challenging aspect of research on multi-agent dynamical systems, formulated as non-cooperative stochastic differential/dynamic games (SDGs) with asymmetric dynamic information structures is the presence of strategic interactions among agents, with each one developing beliefs on others in the absence of shared information. This belief generation process involves what is known as second-guessing phenomenon, which generally entails infinite recursions, thus compounding the difficulty of obtaining (and arriving at) an equilibrium. This difficulty is somewhat alleviated when there is a high population of agents (players), in which case strategic interactions at the level of each agent become much less pronounced. This leads, under some structural constraints, to what is known as mean field games (MFGs), which have been the subject of intense research activity during the last ten years or so.

MFGs constitute a class of non-cooperative stochastic differential games where there is a large number of players or agents who interact with each other through a mean field coupling term—also known as the mass behavior or the macroscopic behavior in statistical physics—included in the individual cost functions and/or each agent’s dynamics generated by a controlled stochastic differential equation, capturing the average behavior of all agents. One of the main research issues in MFGs with no hierarchy in decision making is to study the existence, uniqueness and characterization of Nash equilibria with an infinite population of players under specified information structures and further to study finite-population approximations, that is to explore to what extent an infinite-population Nash equilibrium provides an approximate Nash equilibrium for the finite-population game, and what the relationship is between the level of approximation and the size of the population.

Following a general overview of the difficulties brought about by strategic interactions in finite-population SDGs, the talk will dwell on two classes of MFGs: those characterized by risk sensitive (that is, exponentiated) objective functions (known as risk-sensitive MFGs) and those that have risk-neutral (RN) objective functions but with an additional adversarial driving term in the dynamics (known as robust MFGs). In stochastic optimal control, it is known that risk-sensitive (RS) cost functions lead to a behavior akin to robustness, leading to establishment of a connection between RS control problems and RN minimax ones. The talk will explore to what extent a similar connection holds between RS MFGs and robust MFGs, particularly in the context of linear-quadratic problems, which will allow for closed-form solutions and explicit comparisons between the two in both infinite- and finite-population regimes and with respect to the approximation of Nash equilibria in going from the former to the latter. The talk will conclude with a brief discussion of several extensions of the framework, such as to hierarchical decision structures with a small number of players at the top of the hierarchy (leaders) and an infinite population of agents at the bottom (followers) as well as to games where players make noisy observations.


Tamer BaÅŸar has been with the University of Illinois at Urbana-Champaign since 1981, where he holds the academic positions of Swanlund Endowed Chair; Center for Advanced Study (CAS) Professor of Electrical and Computer Engineering; Professor, Coordinated Science Laboratory; Professor, Information Trust Institute; and Affiliate Professor, Mechanical Science and Engineering. He is also the Director of the Center for Advanced Study. At Illinois, he has also served as Interim Dean of Engineering and Interim Director of the Beckman Institute for Advanced Science and Technology. He is a member of the US National Academy of Engineering; Fellow of IEEE, IFAC, and SIAM; a past president of the IEEE Control Systems Society (CSS), the founding president of the International Society of Dynamic Games (ISDG), and a past president of the American Automatic Control Council (AACC). He has received several awards and recognitions over the years, including the highest awards of IEEE CSS, IFAC, AACC, and ISDG, the IEEE Control Systems Technical Field Award, and a number of international honorary doctorates and professorships, most recently an honorary doctorate from KTH, Sweden. He has over 900 publications in systems, control, communications, optimization, networks, and dynamic games, including books on non-cooperative dynamic game theory, robust control, network security, wireless and communication networks, and stochastic networks. He was Editor-in-Chief of the IFAC Journal Automatica between 2004 and 2014, and is currently editor of several book series. His current research interests include stochastic teams, games, and networks; multi-agent systems and learning; data-driven distributed optimization; epidemics modeling and control over networks; security and trust; energy systems; and cyber-physical systems.

A Decoupling Principle in Stochastic Optimal Control and Its Implications

February 28, 2020 – 11:15 AM
TSRB auditorium

Suman Chakravorty

Texas A&M University


The problem of Stochastic Optimal Control is ubiquitous in Robotics and Control since it is the fundamental formulation for decision-making under uncertainty. The answer to the problem can be computed by solving an associated Dynamic Programming (DP) problem. Unfortunately, the DP paradigm is also synonymous with the infamous “Curse of Dimensionality (COD),” a phrase coined by the discoverer of the Dynamic Programming paradigm, Richard Bellman, nearly 60 years ago, to capture the fact that the computational complexity of solving a DP problem grows exponentially in the dimension of the state space of the problem.

In this talk, we will introduce a newly discovered paradigm in stochastic optimal control, called “Decoupling,” that allows us to separate the design of the open and closed loops of a stochastic optimal control problem with continuous control space. This Decoupled solution allows us to break the COD inherent in DP problems, while remaining near-optimal, to third order, to the true stochastic control. The implications of the Decoupled design are examined in the context of Model Predictive Control (MPC) and Reinforcement Learning (RL). We shall introduce two algorithms, called the Trajectory Optimized Perturbation Feedback Control (T-PFC), and the Decoupled Data based Control(D2C), for the MPC and RL problems respectively. We shall also examine the consequences of the decoupling principle in partially observed/ belief space planning problems and present the Trajectory optimized Linear Quadratic Gaussian (T-LQG) algorithm.


Suman Chakravorty obtained his B.Tech in Mechanical Engineering in 1997 from the Indian Institute of Technology, Madras and his PhD in Aerospace Engineering from the University of Michigan, Ann Arbor in 2004. From August 2004- August 2010, he was an Assistant Professor with the Aerospace Engineering Department at Texas A&M University, College Station and since August 2010, he has been an Associate Professor in the department. Dr. Chakravorty’s broad research interests lie in the estimation and control of stochastic dynamical systems with application to autonomous, distributed robotic mapping and planning, and situational awareness problems. He is a member of AIAA, ASME and IEEE. He is an Associate Editor for the ASME Journal on Dynamical Systems, Measurement and Control and the IEEE Robotics and Automation Letters.

On strategic information transmission in Cyber-Socio-Physical Systems

February 14, 2020 – 11:15 AM
TSRB auditorium

Cedric Langbort

University of Illinois at Urbana Champaign


We consider situations in which better informed agents must decide which message to transmit to a decision-making receiver, so as to influence the decision in their favor.

Such scenarios, some versions of which have been considered earlier in Economics under the umbrellas of ‘cheap talk’ and ‘persuasion theory’ have found renewed relevance in a number of cyber-socio-physical contexts, and have interesting connections to both information and game theory.

Starting with the simplest single transmitter-single receiver setup, we present several variants of such strategic information transmission of increasing complexity, in terms of (1) strategic refinement and rationality of the players, (2) information asymmetries and (3) sender network structure, as well as applications.

The nature of equilibrium sending strategies, as well as the sender’s ability to reach an appropriate decision despite deception vary widely depending on these assumptions, thus illustrating the subtlety of these deception games.


Cedric Langbort is an Associate Professor of Aerospace Engineering at the University of Illinois at Urbana–Champaign (UIUC), where he is also affiliated with the Decision & Control Group at the Coordinated Science Lab (CSL), and the Information Trust Institute. Prior to joining UIUC in 2006, he studied at the Ecole Nationale Superieure de l’Aeronautique et de l’Espace-Supaero in Toulouse (France), the Institut Non-Lineaire in Nice (France), and Cornell University, from which he received the Ph.D. in Theoretical & Applied Mechanics in January 2005. He also spent a year and a half as a postdoctoral scholar in the Center for the Mathematics of Information at Caltech. He works on applications of control, game, and optimization theory to a variety of fields; most recently to “smart infrastructures” problems within the Center for People & Infrastructures which he co-founded and co-directs at CSL. He is a recipient of the NSF CAREER Award, the advisor of an IEEE CDC Best Student Paper Award recipient, and has been an associate editor for OCAM, the journal of Optimal Control Application and Methods, as well as for Systems & Control Letters.

Brockett’s Topological Obstruction

January 22, 2020 – 11:00 AM
TSRB 442

Magnus Egerstedt

Professor and School Chair
Electrical and Computer Engineering


Dr. Magnus Egerstedt is the Steve W. Chaddick School Chair and Professor in the School of Electrical and Computer Engineering at the Georgia Institute of Technology, where he also holds secondary appointments in Mechanical Engineering, Aerospace Engineering, and Interactive Computing. He received the M.S. degree in Engineering Physics and the Ph.D. degree in Applied Mathematics from the Royal Institute of Technology, Stockholm, Sweden, the B.A. degree in Philosophy from Stockholm University, and was a Postdoctoral Scholar at Harvard University.

Dr. Egerstedt conducts research in the areas of control theory and robotics, with particular focus on control and coordination of complex networks, such as multi-robot systems, mobile sensor networks, and cyber-physical systems. Magnus Egerstedt is a Fellow of both IEEE and IFAC, and is a foreign member of the Royal Swedish Academy of Engineering Sciences. He has received a number of teaching and research awards for his work, including the John R. Ragazzini Award from the American Automatic Control Council, the O. Hugo Schuck Best Paper Award from the American Control Conference, the Best Multi-Robot Paper Award from the IEEE International Conference on Robotics and Automation, and the Alumni of the Year Award from the Royal Institute of Technology.

Universal approximation of input-output maps by temporal convolutional nets

November 22, 2019 – 11:15 AM
TSRB Auditorium

Max Raginsky

University of Illinois at Urbana Champaign


A number of problems in machine learning involve sequence-to-sequence transformations, i.e., nonlinear operators that map an input sequence to an output sequence. Traditionally, such input-output maps have been modeled using discrete-time recurrent neural nets. However, there has been a recent shift in sequence-to-sequence modeling from recurrent network architectures to autoregressive convolutional network architectures. These temporal convolutional nets (TCNs) allow for easily parallelizable training and operation, while still achieving competitive performance. In this talk, based on joint work with Joshua Hanson, I will show that TCNs are universal approximators for a large class of causal and time-invariant input-output maps that have approximately finite memory. Specifically, I will present quantitative approximation rates for deep TCNs with rectified linear unit (ReLU) activation functions in terms of the width and depth of the network and the modulus of continuity of the original input-output map. Next, I will show how to apply these results to input-output maps with incrementally stable nonlinear state-space realizations. As an example, I will discuss a class of nonlinear systems of Lur’e type that satisfy a variant of the discrete-time circle criterion.


Maxim Raginsky received the B.S. and M.S. degrees in 2000 and the Ph.D. degree in 2002 from Northwestern University, all in Electrical Engineering. He has held research positions with Northwestern, the University of Illinois at Urbana-Champaign (where he was a Beckman Foundation Fellow from 2004 to 2007), and Duke University. In 2012, he has returned to the UIUC, where he is currently an Associate Professor with the Department of Electrical and Computer Engineering and the Coordinated Science Laboratory. He also holds a courtesy appointment with the Department of Computer Science.

Formal Verification of End-to-End Deep Reinforcement Learning

November 08, 2019 – 11:15 AM
TSRB Auditorium

Yasser Shoukry

University of California, Irvine


From simple logical constructs to complex deep neural network models, Artificial Intelligence (AI)-agents are increasingly controlling physical/mechanical systems. Self-driving cars, drones, and smart cities are just examples of such systems to name a few. However, regardless of the explosion in the use of AI within a multitude of cyber-physical systems (CPS) domains, the safety, and reliability of these AI-enabled CPS is still an understudied problem. Mathematically based techniques for the specification, development, and verification of software and hardware systems, also known as formal methods, hold the promise to provide appropriate rigorous analysis of the reliability and safety of AI-enabled CPS. In this talk, I will discuss our work on applying formal verification techniques to provide formal verification of the safety of autonomous vehicles controlled by end-to-end machine learning models and the synthesis of certifiable end-to-end neural network architectures.


Yasser Shoukry is an Assistant Professor in the Department of Electrical Engineering and Computer Science at the University of California, Irvine where he leads the Resilient Cyber-Physical Systems Lab. Before joining UCI, he spent two years as an assistant professor at the University of Maryland, College Park. He received his Ph.D. in Electrical Engineering from the University of California, Los Angeles in 2015. Between September 2015 and July 2017, Yasser was a joint postdoctoral researcher at UC Berkeley, UCLA, and UPenn. His current research focuses on the design and implementation of resilient cyber-physical systems and IoT. His work in this domain was recognized by the NSF CAREER Award, the Best Demo Award from the International Conference on Information Processing in Sensor Networks (IPSN) in 2017, the Best Paper Award from the International Conference on Cyber-Physical Systems (ICCPS) in 2016, and the Distinguished Dissertation Award from UCLA EE department in 2016. In 2015, he led the UCLA/Caltech/CMU team to win the NSF Early Career Investigators (NSF-ECI) research challenge. His team represented the NSF- ECI in the NIST Global Cities Technology Challenge, an initiative designed to advance the deployment of Internet of Things (IoT) technologies within a smart city. He is also the recipient of the 2019 George Corcoran Memorial Award for his contributions to teaching and educational leadership in the field of CPS and IoT.