Plenary Speakers
-
Domitilla del Vecchio (MIT)
-
Emilio Frazzoli (ETH/nuTonomy)
-
Fredrik Gustafsson (Linkoeping)
-
Maurice Heemels (Eindhoven)
-
Mihailo Jovanovic (USC)
-
Naomi Leonard (Princeton)
-
Ben Recht (Berkeley)
-
Sri Sarma (Hopkins)
-
Alireza Tahbaz-Salehi (Northwestern)
- Dawn Tilbury (Michigan/NSF)
Talks
Engineering Biological Networks: A Control Theoretic
Framework
Biological networks control every aspect of life and the
possibility to engineer them de-novo has opened the path to
a number of exciting applications, from revolutionary drugs
and clean energy, to programmed bugs that track and kill
cancer cells. A promising approach to engineer a complex
system is to compose simpler modules together, under the
assumption that these modules will keep their input/output
behavior unchanged upon insertion into a larger system.
Indeed, this is a pervasive approach used today to engineer
biological circuits in the field of synthetic biology.
However, when engineered biological modules are composed
together, their input/output salient properties often change
in rather astonishing ways. This is typically handled by
re-designing every component of a system once a new
component is added, leading to a daunting, combinatorial,
optimization problem. This is a current bottleneck to
the ability of engineering biological networks that
behave as predicted and are sufficiently sophisticated to
tackle real-life problems. In this talk, I will give an
overview of a number of causes of modularity failing in
engineered biological networks, focusing on problems of
loading and resource sharing, the latter being a current
major issue in the field. I will introduce a control
theoretic framework, grounded on the concept of
retroactivity, that has equipped us with both a system-level
way to model lack of modularity and control theoretic
solutions to restore it. I will conclude the talk by
describing how concepts from controls and networks may be
adapted to address the new physical constraints that biology
imposes.
|
Domitilla Del Vecchio received
the Ph. D. degree in Control and Dynamical Systems
from the California Institute of Technology,
Pasadena, and the Laurea degree in Electrical
Engineering (Automation) from the University of Rome
at Tor Vergata in 2005 and 1999, respectively. From
2006 to 2010, she was an Assistant Professor in the
Department of Electrical Engineering and Computer
Science and in the Center for Computational Medicine
and Bioinformatics at the University of Michigan,
Ann Arbor. In 2010, she joined Department of
Mechanical Engineering at the Massachusetts
Institute of Technology (MIT), where she is
currently Professor and member of the Synthetic
Biology Center. She is a recipient of the 2016 Bose
Research Award (MIT), the Donald P. Eckman Award
from the American Automatic Control Council (2010),
the NSF Career Award (2007), the American Control
Conference Best Student Paper Award (2004), and the
Bank of Italy Fellowship (2000).
|
Autonomy and Decision Making in Ride Hailing Networks:
Progress and Open Questions
In recent years, many of the most urgent questions related
to autonomous vehicles were purely technological and vehicle
centered: how safe are these vehicles, how comfortable are
they, what scenarios can they handle? As public trials
worldwide are becoming more common, capabilities and
limitations are becoming clearer. On the other hand,
much less systematic research is available on the operation
and impact of large-scale fleets of autonomous vehicles
offering mobility-on-demand services. What are the key
parameters and constraints on operation? What will be the
impact on congestion, parking, mass transit? How many
seats should the robotic taxis have? Answering these
questions is an exercise that requires not only technical
know-how but also the consideration of societal and economic
aspects. While much is still in the dark, research of most
recent years has answered some of the open questions. This
talk will provide insights into these answers of what we
already know about autonomous ride-hailing networks.
Moreover, it will also highlight some of the aspects which
until today remain unanswered.
|
Emilio Frazzoli is a
professor of Dynamic Systems and Control at ETH
Zurich, and Chief Scientist of Aptiv's Autonomous
Mobility group. His main research interest are in
robotics, autonomous systems, and intelligent
mobility. In acknowledgement of his seminal work in
these fields, he has received several awards,
including the the 2015 IEEE George S.
Axelby Award and the 2017 IEEE Kiyo Tomiyasu
Award, and has been named an IEEE Fellow in 2019. He
has published more than 200 papers in the fields of
robotics, autonomous vehicles, and drones. A
former full professor at MIT, he directed the
research group that first demonstrated an autonomous
mobility ("robotaxi") service to the public, and
performed the first analysis of the social and
economic impact of such a service, based on real
transportation data. In 2013 he founded nuTonomy
with Karl Iagnemma, and served as its Chief
Technology Officer until its acquisition by
Aptiv in 2017. |
Distributed fusion with communication constraints
Fusion in this context refers to combining estimates in
local nodes to a global estimate. From a statistical point
of view, the first (mean vector) and second (covariance
matrix) order moments can easily be combined if the
estimates are independent and the full covariance matrix can
be communicated. However, if either of this is not true,
approximate safe fusion strategies must be applied. Safe
fusion should yield a conservative covariance matrix which
must not be smaller than the true one. We survey some known
strategies in literature, and present some new ones for the
case when only the diagonal of the covariance matrix is
communicated from the local nodes. We describe a real world
scenario where thousands of taxi cars compute local
estimates of the local road condition, which is combined to
a global road condition map.
|
Fredrik Gustafsson is
professor in Sensor Informatics at Department of
Electrical Engineering, Linkoping University, since
2005. He received the M.Sc. degree in electrical
engineering 1988 and the Ph.D. degree in Automatic
Control, 1992, both from Linkoping University.
He is the author of five books, more than 200
conference papers, 100 journal papers and some 30
patents. His current h-index is 58 (Google Scholar).
He has supervised 25 PhD and more than 200 master
theses. He was an associate editor for IEEE
Transactions of Signal Processing 2000-2006, IEEE
Transactions on Aerospace and Electronic Systems
2010-2012, and EURASIP Journal on Applied Signal
Processing 2007-2012. He was awarded the Arnberg
prize by the Royal Swedish Academy of Science (KVA)
2004, elected member of the Royal Academy of
Engineering Sciences (IVA) 2007, and elevated to
IEEE Fellow 2011. In 2014, he was awarded
a Distinguished Professor grant from the
Swedish Research Council. He was an adjunct
entrepreneurial professor at Twente University
2012-2013.
He was awarded the Harry Rowe Mimno Award 2011 for
the tutorial "Particle Filter Theory and Practice
with Positioning Applications", which was published
in the AESS Magazine in July 2010. This paper was
ranked #4 out of 13077 articles published in the
category IEEE Engineering Aerospace between 2010 and
2014 based on Web of Science citation data. He was a
co-author of "Smoothed state estimates under abrupt
changes using sum-of-norms regularization" that
received the Automatica paper prize in 2014. He is a
co-founder of the companies NIRA Dynamics
(automotive safety, including tire pressure
monitoring systems found in about 50 million cars
today), Softube (plug-ins used in tens of thousands
of music studios), and Senion (indoor navigation for
smartphones deployed in more than 30 countries in
all six continents).
|
Event-triggered communication and control in networked
systems
Computer and communication technologies are developing at
tremendous speeds leading to an increasingly networked and
wireless world. These technologies bring new opportunities
for the control of (large-scale) interconnected systems, but
also raise new challenging questions for the resulting
networked control systems in which more and more data is
exchanged. One important question is how to determine the
time instances at which data is exchanged between
(sub)systems, agents, sensors, controllers and/or actuators
such that desirable closed-loop stability and performance
properties are guaranteed, while keeping the number of
transmissions small. To efficiently use the available
communication resources, we move away from the classical
time-triggered control and communication schemes in which
transmissions take place periodically in time. Instead, we
propose to use event-triggered schemes as alternative and
(more) resource-aware paradigms, as it seems natural to
trigger control and communication actions by well-designed
events involving the system's state, output or any other
locally available information. We will discuss the main
ideas, recent advances, and open questions for these
event-triggered control and communication schemes in the
context of distributed and multi-agent systems. The focus
will be on periodic event-triggered control (PETC) schemes
that unite the benefits of time-triggered control regarding
ease of digital implementation on the one hand and
event-triggered control to reduce communication on the other
hand. The theory will be complemented by experiments in the
area of cooperative driving (platooning) and robotics
exploiting wireless communication.
|
Maurice Heemels received the
M.Sc. degree in mathematics and the Ph.D. degree in
control theory (both summa cum laude) from the
Eindhoven University of Technology (TU/e), the
Netherlands, in 1995 and 1999, respectively. From
2000 to 2004, he was with the Electrical Engineering
Department, TU/e and from 2004 to 2006 with the
Embedded Systems Institute (ESI). Since 2006, he has
been with the Department of Mechanical Engineering,
TU/e, where he is currently a Full Professor. He
held visiting professor positions at the Swiss
Federal Institute of Technology (ETH), Switzerland
(2001) and at the University of California at Santa
Barbara (2008). In 2004, he worked also at the
company Oce, the Netherlands. His current research
interests include hybrid and cyber-physical systems,
networked and event-triggered control systems and
constrained systems including model predictive
control. Maurice served/s on the editorial boards of
Automatica, Nonlinear Analysis: Hybrid Systems,
Annual Reviews in Control, and IEEE Transactions on
Automatic Control. He was a recipient of a personal
VICI grant (1.5 MEuro) awarded by STW (Dutch
Technology Foundation, NWO/TTW) and is the current
chair of the IFAC Technical Committee on Networked
Systems. He is the founding father of a bi-annual
PhD school on multi-disciplinary research topics
such as hybrid, networked and cyber-physical systems
educating over 600 PhD students worldwide since
2003. He was/is IPC (co-)chair of IFAC ADHS'12,
ECC'13, IFAC NECSYS'13, IFAC ADHS'18, and ECC’21. He
is a Fellow of the IEEE.
|
Non-smooth composite optimization: A primal-dual method
based on the proximal augmented Lagrangian
Several problems in the design of large-scale networks of
dynamical systems can be formulated as non-smooth composite
optimization problems in which the objective function is
given by the sum of a differentiable term and a
non-differentiable regularizer. For example, the edge
addition to improve network performance, redesign of the
edge weights to optimize certain performance metric, and
selection of important nodes in a network with a given
topology fit into this category. In this talk, we will
describe a primal-dual method based on the proximal
augmented Lagrangian for solving this class of non-smooth
optimization problems. After introducing an auxiliary
variable, we utilize the proximal operator of the nonsmooth
regularizer to transform the associated augmented Lagrangian
into a function that is once, but not twice, continuously
differentiable. Saddle points of this function, which we
call proximal augmented Lagrangian, correspond to the
solution of the original optimization problem. This function
is used to develop customized algorithms based on the first
and second order primal-dual methods. When the
differentiable component of the objective function is
strongly convex with a Lipschitz continuous gradient, we
employ the theory of integral quadratic constraints to prove
global exponential stability of the primal-descent
dual-ascent gradient method. We also use a generalization of
the Hessian to define second order updates on this function
and prove global exponential stability of the corresponding
differential inclusion. We close the talk by discussing the
classes of problems that are amenable to distributed
optimization and compare performance of the developed method
to the state-of-the-art alternatives.
|
Mihailo R. Jovanovic is a
professor in the Ming Hsieh Department of Electrical
and Computer Engineering and the founding director
of the Center for Systems and Control at the
University of Southern California. He was a faculty
in the Department of Electrical and Computer
Engineering at the University of Minnesota,
Minneapolis, from December 2004 until January 2017,
and has held visiting positions with Stanford
University and the Institute for Mathematics and its
Applications. His current research focuses on
large-scale and distributed optimization, design of
controller architectures, dynamics and control of
fluid flows, and fundamental performance limitations
in the control of large-scale networks of dynamical
systems. He serves as an Associate Editor of the
IEEE Transactions on Control of Network Systems, and
had served as the Chair of the APS External Affairs
Committee, a Program Vice-Chair of the 55th IEEE
Conference on Decision and Control, an Associate
Editor of the SIAM Journal on Control and
Optimization (from 2014 until 2017), and an
Associate Editor of the IEEE Control Systems Society
Conference Editorial Board (from 2006 until 2010).
Prof. Jovanovic is a fellow of APS and IEEE. He
received a CAREER Award from the National Science
Foundation in 2007, the George S. Axelby Outstanding
Paper Award from the IEEE Control Systems Society in
2013, and the Distinguished Alumnus Award from the
UC Santa Barbara in 2014. |
Multi-Armed Bandits and Distributed Decision Making in
Networks
Decision making in explore-exploit tasks, from resource
allocation to search in an uncertain environment, can be
modeled using multi-armed bandit (MAB) problems, where the
decision maker chooses sequentially in time among multiple
options with uncertainty in rewards. I will present a
framework for evaluating and designing distributed decision
making in heterogeneous multi-player multi-armed bandit
problems in which agents communicate according to a network
graph. For agents sharing their estimates with
neighbors, we propose an explore-exploit centrality index
that distinguishes an agent’s performance as a function of
its location in the network. For agents only observing
their neighbors’ choices and rewards, we propose an index
that distinguishes performance as a function of the agent’s
“sociability” and that its neighbors. I will discuss
applications to robotic teams and to human groups.
|
Naomi Ehrich Leonard is Edwin
S. Wilsey Professor of Mechanical and Aerospace
Engineering and associated faculty in Applied and
Computational Mathematics at Princeton
University. She is a MacArthur Fellow, and
Fellow of the American Academy of Arts and Sciences,
SIAM, IEEE, IFAC, and ASME. She received her
BSE in Mechanical Engineering from Princeton
University and her PhD in Electrical Engineering
from the University of Maryland. Her research
is in control and dynamics with application to
multi-agent systems, autonomous vehicles, robotic
networks, collective animal behavior, and human
decision dynamics.
|
The Merits of Models in Continuous Reinforcement Learning
Classical control theory and machine learning have similar
goals: acquire data about the environment, perform a
prediction, and use that prediction to impact the world.
However, the approaches they use are frequently at odds.
Controls is the theory of designing complex actions from
well-specified models, while machine learning makes
intricate, model-free predictions from data alone. For
contemporary autonomous systems, some sort of hybrid may be
essential in order to fuse and process the vast amounts of
sensor data recorded into timely, agile, and safe decisions.
In this talk, I will examine the relative merits of
control-centric model-based and learning-centric model-free
methods in data-driven control problems. I will discuss
quantitative estimates on the number of measurements
required to achieve a high quality control performance and
statistical techniques that can distinguish the relative
power of different methods. In particular, I will show how
model-free methods are considerably less sample efficient
than their model-based counterparts. I will also describe
how notions of robustness, safety, constraint satisfaction,
and exploration can be transparently incorporated in
model-based methods. I will conclude with a discussion of
possible positive roles for model-free methods in
contemporary autonomous systems that may mitigate their high
sample complexity and lack of reliability and versatility.
|
Benjamin Recht is an Associate
Professor in the Department of Electrical
Engineering and Computer Sciences at the University
of California, Berkeley. Ben's research group
studies the theory and practice of optimization
algorithms with a particular focus on applications
in machine learning and control. Ben is the
recipient of a Presidential Early Career Award for
Scientists and Engineers, an Alfred P. Sloan
Research Fellowship, the 2012 SIAM/MOS Lagrange
Prize in Continuous Optimization, the 2014 Jamon
Prize, the 2015 William O. Baker Award for
Initiatives in Research, and the 2017 NIPS Test of
Time Award.
|
Risk-taking bias in human decision-making is controlled
via a distributed right-left brain push-pull network
A person's decisions vary even when options stay the same,
like when a gambler changes bets despite constant odds of
winning. Internal bias (e.g., emotion) contributes to
this variability and is shaped by past outcomes, yet its
neurobiology during decision-making is not well understood.
To map neural networks encoding bias, we administered a
gambling task to 10 participants implanted with
intracerebral depth electrodes in cortical and subcortical
structures. We predicted the variability in betting behavior
within and across patients by individual bias, which is
estimated through a dynamical model of choice. Our analysis
further revealed that high-frequency activity increased in
the right hemisphere when participants were biased towards
risky bets, while it increased in the left hemisphere when
participants were biased away from risky bets. Our findings
provide the first electrophysiological evidence that
risk-taking bias is a lateralized distributed push-pull
neural system governing counterintuitive and highly variable
decision-making in humans.
|
Sri Sarma received the B.S.
degree in electrical engineering from Cornell
University, Ithaca NY, in 1994; and an M.S.
and Ph.D. degrees in Electrical Engineering
and Computer Science from Massachusetts Institute of
Technology in, Cambridge MA, in 1997 and 2006,
respectively. From 2000-2003 she took a leave of
absence to start a data analytics company. From
2006--2009, she was a Postdoctoral Fellow in the
Brain and Cognitive Sciences Department at the
Massachusetts Institute of Technology,
Cambridge. She is now an associate professor
in the Institute for Computational Medicine,
Department of Biomedical Engineering, at Johns
Hopkins University, Baltimore MD. Her research
interests include modeling, estimation and control
of neural systems using electrical stimulation. She
is a recipient of the GE faculty for the future
scholarship, a National Science Foundation graduate
research fellow, a L'Oreal For Women in Science
fellow, the Burroughs Wellcome Fund Careers at the
Scientific Interface Award, the Krishna Kumar New
Investigator Award from the North American
Neuromodulation Society, and a recipient of the
Presidential Early Career Award for Scientists and
Engineers (PECASE) and the Whiting School of
Engineering Robert B. Pond Excellence in Teaching
Award.
|
Networks and the Macroeconomy
This talk gives an overview of the recent literature on
production networks in macroeconomics. It starts by
presenting the theoretical foundations of how input-output
linkages can function as a mechanism for the propagation and
amplification of shocks and how networks can transform
microeconomic shocks into macroeconomic fluctuations. It
then provides a brief guide to the growing literature that
explores these themes empirically and quantitatively.
|
Alireza Tahbaz-Salehi is an
associate professor of Managerial Economics and
Decision Sciences Department at the Kellogg School
of Management, Northwestern University. Prior to
joining Kellogg, he was the Daniel W. Stanton
Associate Professor of Business at Columbia Business
School. His research focuses on the implications of
network economies for information aggregation,
business cycle fluctuations, and financial
stability. |
A Multi-agent Distributed Control Approach to Complex
Manufacturing Systems
Many manufacturing systems today have been optimized for
mass production, making the same products over and over
again with high quality and low cost. There is an increased
demand for customized or even personalized production, which
can be possible while utilizing many of the same machines
currently existing on plant floors. However, the
control systems must be completely redefined. The
Internet of Things and networked control systems are key
enabling technologies to realize this vision.
Decentralized control strategies, specifically agent-based
control, can be used to enable customized and personalized
production, while improving the flexibility and
responsiveness of manufacturing systems. A
manufacturing plant floor has both resource agents,
representing the processing and material handling resources
available, as well as product agents, representing the parts
that traverse through the factory, being transformed from
raw materials to finished products. Agents each have
their own goals, and make decisions based on these goals,
their communications with each other, and information
available from the physical system. This presentation
will cover our recent work on product agents, and discuss
how this decentralized approach can lead to improved
productivity. Implementation on a small-scale
automated testbed will be presented.
|
Dawn M. Tilbury received the
B.S. degree in Electrical Engineering, summa cum
laude, from the University of Minnesota in 1989, and
the M.S. and Ph.D. degrees in Electrical Engineering
and Computer Sciences from the University of
California, Berkeley, in 1992 and 1994,
respectively. In 1995, she joined the faculty
of the University of Michigan, Ann Arbor, where she
is currently Professor of Mechanical Engineering
with a joint appointment in Electrical Engineering
and Computer Science. Her research interests
lie broadly in the area of control systems,
including applications to robotics and manufacturing
systems. She has published more than 150 articles in
refereed journals and conference proceedings.
She was elected Fellow of the IEEE in 2008 and
Fellow of the ASME in 2012, and is a Life Member of
SWE. Since June of 2017, she has been the
Assistant Director for Engineering at the National
Science Foundation.
|
|