seminar

Design of Robotics and Embedded systems, Analysis, and Modeling Seminar (DREAMS)

Fall 2016

The Design of Robotics and Embedded systems, Analysis, and Modeling Seminar (DREAMS) occurs weekly on Mondays from 4.10-5.00 p.m. in 250 Sutardja Dai Hall.

Similar to last year, DREAMS has joined forces with the Control Theory Seminar and the CITRIS People and Robots Seminar CPAR.

The Design of Robotics and Embedded systems, Analysis, and Modeling Seminar topics are announced to the DREAMS list, which includes the chessworkshop workgroup, which includes the chesslocal workgroup.

Information on the seminar series might be useful for potential speakers. If you have any questions about DREAMS, please contact Dorsa Sadigh. If you want to subscribe to our mailing list, please drop me a line.

Seminars from previous semesters can be found here.

Today's Seminar Talk

Schedule

Alessandro Abate August 22, 2016
Mykel Kochenderfer August 29, 2016
Mark Mueller September 12, 2016
Joao Hespanha September 19, 2016
Tom Henzinger September 22, 2016
Meeko Oishi September 26, 2016
Arthur Krener September 29, 2016
Marilena Vendittelli October 04, 2016 UPCOMING
Aws Albarghouthi October 10, 2016 UPCOMING
Tom Griffiths October 17, 2016 UPCOMING
Nancy Amato October 24, 2016 UPCOMING
Sergey Levine October 31, 2016 UPCOMING
Ian Mitchell November 07, 2016 UPCOMING
Marco Pavone November 14, 2016 UPCOMING
Hamsa Balakrishnan November 21, 2016 UPCOMING
Karl Johansson November 28, 2016 UPCOMING
Yon Visell December 05, 2016 UPCOMING

Data-driven and model-based quantitative verification and correct-by-design synthesis of CPS

Aug 22, 2016, 4-5pm, 250 SDH, Alessandro Abate, University of Oxford.

Slides

Abstract

I discuss a new and formal, measurement-driven and model-based automated verification and synthesis technique, to be applied on quantitative properties over systems with partly unknown dynamics. I focus on physical systems (with spatially continuous variables, possibly noisy), driven by external inputs and accessed under noisy measurements, and suggest that the approach can be as well generalised over CPS. I formulate this new setup as a data-driven Bayesian model inference problem, formally embedded within a quantitative, model-based verification procedure.

While emphasising the generality of the approach over a number of diverse model classes, this talk zooms in on systems represented via stochastic hybrid models (SHS), which are probabilistic models with heterogeneous dynamics (continuous/discrete, i.e. hybrid, as well as nonlinear) - as such, SHS are quite a natural framework for CPS. With focus on model-based verification procedures, I provide the characterisation of general temporal specifications based on Bellman’s dynamic programming. The computation of such properties and the synthesis of related control architectures optimising properties of interest, is attained via the development of abstraction techniques based on quantitative approximations. Theory is complemented by algorithms, all packaged in a software tool (FAUST^2) that is freely available to users.

Bio:

Alessandro Abate is an Associate Professor in the Department of Computer Science at the University of Oxford, and a Fellow of the Alan Turing Institute in London (UK). He received a Laurea in Electrical Engineering in October 2002 from the University of Padova (IT), an MS in May 2004 and a PhD in December 2007, both in Electrical Engineering and Computer Sciences, at UC Berkeley (USA). He has been an International Fellow in the CS Lab at SRI International in Menlo Park (USA), and a PostDoctoral Researcher at Stanford University (USA), in the Department of Aeronautics and Astronautics. From June 2009 to mid 2013 he has been an Assistant Professor at the Delft Center for Systems and Control, TU Delft (NL).

His research interests are in the formal verification and control of complex (probabilistic and hybrid) models, in the integration of data-based learning aspects within these deductive techniques, and in the application of complex models over a number of domains, particularly in energy and in systems biology.


Building Trust in Decision Support Systems for Aerospace

Aug 29, 2016, 4-5pm, 250 SDH, Mykel Kochenderfer, Stanford University.

Slides

Abstract

Starting in the 1970s, decades of effort went into building human-designed rules for providing automatic maneuver guidance to pilots to avoid mid-air collisions. The resulting system was later mandated worldwide on all large aircraft and significantly improved the safety of the airspace. Recent work has investigated the feasibility of using partially observable Markov decision processes (POMDPs) and advanced computational techniques to derive optimized decision logic that better handles various sources of uncertainty and balances competing system objectives. This approach has resulted in a system called Airborne Collision Avoidance System (ACAS) X that significantly reduces the risk of mid-air collision while also reducing the alert rate, and it is in the process of becoming the next international standard. Using ACAS X as a case study, this talk will discuss lessons learned about building trust in advanced decision support systems. This talk will also outline research challenges in facilitating greater levels of automation and integrating unmanned aircraft into the airspace.

Bio:

Mykel Kochenderfer is Assistant Professor of Aeronautics and Astronautics at Stanford University. Prior to joining the faculty, he was at MIT Lincoln Laboratory where he worked on airspace modeling and aircraft collision avoidance, with his early work leading to the establishment of the ACAS X program. He received a Ph.D. from the University of Edinburgh and B.S. and M.S. degrees in computer science from Stanford University. Prof. Kochenderfer is the director of the Stanford Intelligent Systems Laboratory (SISL), conducting research on advanced algorithms and analytical methods for the design of robust decision making systems. He is the author of "Decision Making under Uncertainty: Theory and Application" from MIT Press. He is a third generation pilot.


Multicopter dynamics and control: surviving the complete loss of multiple actuators and quickly generating trajectories

Sep 12, 2016, 4-5pm, 250 SDH, Mark Mueller, University of California, Berkeley.

Slides

Abstract

Multicopters are increasingly becoming part of our everyday lives, with current and future applications including delivery services, entertainment, and aerial sensing. These systems are expected to be safe and to have a high degree of autonomy. This talk will look at the dynamics of a multicopter, and then discuss multicopter failsafe strategies and fast trajectory generation.

The first part of the talk presents an actuator redundancy scheme for multicopters, allowing e.g. a quadcopter to maintain controlled flight despite the complete failure of half its actuators. This redundancy is a crucial part of the safety concept of a NYC Broadway musical drone performance, where eight quadcopters perform eight shows per week in front of an audience of 2'000 people, without any protective nets. The related 'monospinner' will also be presented: with only one moving part, it is potentially the world's mechanically simplest, controllable flying vehicle.
In the second part, a computationally light-weight strategy for generating quadrocopter motion primitives is presented. This trajectory generation can evaluate and compare on the order of one million motion primitives per second on a standard laptop computer. These motion primitives are designed to be fast to compute and verify (at the expense of optimality), while being flexible with respect to initial and final states. This allows to encode highly dynamic tasks with complicated end goals, such as catching a thrown ball.

Bio:

Mark W. Mueller completed his PhD studies, advised by Prof. Raffaello D'Andrea, at the Institute for Dynamic Systems and Control at the ETH Zurich at the end of 2015. He received a bachelors degree from the University of Pretoria, and a masters from the ETH Zurich in 2011, both in Mechanical Engineering. After working a brief period at a startup company, he joined the Mechanical Engineering Department at Berkeley in August of this year.


Opportunities and Challenges in Control Systems arising from Ubiquitous Communication and Computation

Sep 19, 2016, 4-5pm, 250 SDH, Joao Hespanha, University of California, Santa Barbara.

Slides

Abstract

Advances in VLSI (Very Large Scale Integration) design and fabrication have resulted in the availability of low-cost, low-power, small-sized devices that have significant computational power and are able to communicate wirelessly. In addition, advances in MEMS (Micro Electric Mechanical Systems) technology have resulted in wide availability of solid-state sensors and actuators. The net result is ubiquitous sensing, communication, and computation that can be incorporated into small low-power devices.

In this talk, I will discuss how the above-mentioned technological advances present important opportunities and interesting challenges for control system designers. To this effect, I will discuss how the introduction of digital communication in control loops gives rise to a need for new tools for the design and analysis of feedback control systems. I will also describe recent work demonstrating that feedback control based on on-line optimization is a viable approach to solve a wide range of control problem.

Bio:

João P. Hespanha received his Ph.D. degree in electrical engineering and applied science from Yale University, New Haven, Connecticut in 1998. From 1999 to 2001, he was Assistant Professor at the University of Southern California, Los Angeles. He moved to the University of California, Santa Barbara in 2002, where he currently holds a Professor position with the Department of Electrical and Computer Engineering. Prof. Hespanha is the Chair of the Department of Electrical and Computer Engineering and a member of the Executive Committee for the Institute for Collaborative Biotechnologies (ICB). Dr. Hespanha is a Fellow of the IEEE and was an IEEE distinguished lecturer from 2007 to 2013. His current research interests include hybrid and switched systems; multi-agent control systems; distributed control over communication networks (also known as networked control systems); the use of vision in feedback control; stochastic modeling in biology; and network security.


The Quest for Average Response Time

Sep 22, 2016, 4-5pm, 250 SDH, Tom Henzinger, IST Austria.

Slides

Abstract

Responsiveness -the requirement that every request to a system be eventually handled- is one of the fundamental liveness properties of a reactive system and lies at the heart of all methods for specifying and verifying liveness. Average response time is a quantitative measure for the responsiveness requirement used commonly in performance evaluation. The static computation of average response time has proved remarkably elusive even for finite-state models of reactive systems. We present, for the first time, a robust formalism that allows the specification and computation of quantitative temporal properties including average response time. The formalism is based on nested weighted automata, which can serve as monitors for measuring the response time of a reactive system. We show that the average response time can be computed in exponential space for nondeterministic finite-state models of reactive systems and in polynomial time for probabilistic finite-state models. This work is joint with Krishnendu Chatterjee and Jan Otop.

Bio:

Thomas A. Henzinger is president of IST Austria (Institute of Science and Technology Austria). He holds a Dipl.-Ing. degree in Computer Science from Kepler University in Linz, Austria, an M.S. degree in Computer and Information Sciences from the University of Delaware, a Ph.D. degree in Computer Science from Stanford University (1991), and a Dr.h.c. from Fourier University in Grenoble, France (2012) and from Masaryk University in Brno, Czech Republic (2015). He was Assistant Professor of Computer Science at Cornell University (1992-95), Assistant Professor (1996-97), Associate Professor (1997-98), and Professor (1998-2004) of Electrical Engineering and Computer Sciences at the University of California, Berkeley. He was also Director at the Max-Planck Institute for Computer Science in Saarbruecken, Germany (1999) and Professor of Computer and Communication Sciences at EPFL in Lausanne, Switzerland (2004-09). His research focuses on modern systems theory, especially models, algorithms, and tools for the design and verification of reliable software, hardware, and embedded systems. His HyTech tool was the first model checker for mixed discrete-continuous systems. He is an ISI highly cited researcher, a member of Academia Europaea, a member of the German Academy of Sciences (Leopoldina), a member of the Austrian Academy of Sciences, a Fellow of the AAAS, a Fellow of the ACM, and a Fellow of the IEEE. He has received the Milner Award of the Royal Society, the Wittgenstein Award of the Austrian Science Fund, and an ERC Advanced Investigator Grant.


Collaborative control for human-in-the-loop systems: Optimal interface design and reachability-based collaborative navigation

Sep 26, 2016, 4-5pm, 250 SDH, Meeko Oishi, University of New Mexico.

Slides

Abstract

Methods for the analysis and design of human-in-the-loop systems must account for interactions between the automation, the human, and the environment. We consider two problems: 1) user-interface design, and 2) collaborative navigation in dynamic, uncertain environments. The user-interface, which provides information to the user about the underlying automation, and allows the user to issue input commands to the system, is key for enabling situational awareness and trust of the automation, yet is often designed in an ad-hoc fashion. We use sensor placement techniques to determine the optimal elements for display in the user-interface, and exploit submodularity properties to facilitate solution of the resulting combinatorial optimization problem. We additionally consider the problem of collaborative navigation in dynamic, uncertain environments. While assurances of safety are computationally intractable, solutions that exploit the forward reachable set are real-time compatible. We describe a method to compute the forward stochastic reachable set and its probability measure efficiently, that enables robust performance in difficult planning problems.

Bio:

Meeko Oishi is an Associate Professor of Electrical and Computer Engineering at the University of New Mexico. She received the Ph.D. (2004) and M.S. (2000) in Mechanical Engineering from Stanford University, and a B.S.E. in Mechanical Engineering from Princeton University (1998). Her research interests include nonlinear dynamical systems, hybrid control theory, control of human-in-the-loop systems, reachability analysis, and modeling of motor performance and control in Parkinson’s disease. She previously held a faculty position at the University of British Columbia at Vancouver. She is the recipient of the NSF CAREER Award, the UNM Regents’ Lecturer Award, the UNM Teaching Fellowship, the Peter Wall Institute Early Career Scholar Award, the Truman Postdoctoral Fellowship in National Security Science and Engineering, and the George Bienkowski Memorial Prize, Princeton University. She was a Summer Faculty Fellow at AFRL Space Vehicles Directorate 2013–2015.


Adaptive Horizon Model Predictive Control

Sep 29, 2016, 4-5pm, 250 SDH, Arthur Krener, University of California, Davis.

Slides

Abstract

Adaptive Horizon Model Predictive Control (AHMPC) is a scheme for varying the horizon length of Model Predictive Control (MPC) as needed. Its goal is to achieve stabilization with horizons as small as possible so that it can be used on faster or more complicated dynamic process. Beside the standard requirements of MPC including a terminal cost that is a control Lyapunov function, AHMPC requires a terminal feedback that turns the control Lyapunov function into a standard Lyapunov function in some domain around the operating point. But this domain need not be known explicitly. Just as MPC does not compute off-line the optimal cost and the optimal feedback over a large domain instead it computes these quantities on-line when and where they are needed, AHMPC does not compute off-line the domain on which the terminal cost is a control Lyapunov function instead it computes on-line when a state is in this domain.

AHMPC verifies in real time that stabilization is being achieved. This is particularly important when dealing with nonlinear systems because the nonlinear programs that MPC passes to the nonlinear program solver are typically nonconvex. Therefore the solver may return local rather than global solutions and these may fail to be stabilizing. AHMPC detects when the solutions are not stabilizing. If so there is a need to pass different initial guesses to the solver to find global solutions or at least local solutions that are stabilizing.

Bio:

Arthur Krener is a mathematician whose research interests are in developing methods for the control and estimation of nonlinear dynamical systems and stochastic processes. In 1971 he received the PhD in Mathematics from the University of California, Berkeley and joined the faculty of the University of California, Davis. He retired from UCD in 2006 as a Distinguished Professor of Mathematics and he currently is a Distinguished Visiting Professor at the Naval Postgraduate School. He has also held visiting positions at Harvard University, Imperial College, NASA Ames Research Center, the University of California, Berkeley, the University of Paris, the University of Maryland, the University of Padua and North Carolina State University. He is a member of the American Mathematical Association, a Fellow of the Society for Industrial and Applied Mathematics and a Life Fellow of the Institute of Electrical and Electronics Engineers. Krener has held a variety of administrative posts, including Chair of the Department of Mathematics, UC Davis, member of the Committee on Academic Personnel, UC Davis and Chair of the SIAM Activity Group on Control and Systems Theory.


Interaction force reconstruction for humanoid robots

Oct 04, 2016, 3-4pm, 730 SDH, Marilena Vendittelli, Sapienza University of Rome.

Slides

Abstract

Humanoids are, by definition, robotic systems for which the control of interaction forces with the environment is elemental for the accomplishment of any loco-manipulation task. Any feedback controller of the interaction forces would invariably require some form of measure or estimation of the forces actually exchanged between the robot and the environment. In this talk we will discuss the peculiarity of the interaction force estimation problem in the case of humanoids and we will propose methods that do not make use of force sensors and do not limit the interaction to specific points of the humanoid body. The illustration of the proposed approach is completed by experimental validation on the robot NAO. The talk is based on the work in [1] and [2] where two different approaches to interaction force estimation are considered. In [1], an accurate reconstruction of contact forces and identification of the contact point is obtained using joint position and motor current measurements under the assumption of static equilibrium. The method is designed for manipulation task or multi-contact locomotion, where accurate estimation of the contact forces is important for task accomplishment, including equilibrium preservation. Interaction forces are, instead, perceived in [2] through a measure of the equilibrium perturbation. The reconstructed force is used as a guiding force in collaborative tasks and the humanoid locomotion results from the proposed equilibrium preservation strategy.

BIBLIO [1] T. Mattioli, M. Vendittelli, "Interaction force reconstruction for humanoid robots," IEEE Robotics and Automation Letters, 2016. DOI: 10.1109/LRA.2016.2601345

[2] M. Bellaccini, A. Paolillo, L. Lanari, M. Vendittelli, "Manual guidance of humanoid robots without force sensors: preliminary experiments with NAO," 2014 IEEE International Conference on Robotics and Automation, Hong Kong, China, May 2014. DOI: 10.1109/ICRA.2014.6907003

Bio:

Marilena Vendittelli is Assistant Professor at Sapienza University of Rome, Department of Computer, Control and Management Engineering (DIAG), since 2001 and a member of the Robotics Laboratory at DIAG. She worked at LAAS-CNRS in Toulouse, France, first as a visiting scholar in 1995-96 and then as a post-doc in 1997-98, funded by a Marie Curie Research Training Grant. The research activity was focused on planning and control of general (i.e., not transformable in a canonical form) nonholonomic systems. She received the PhD degree in Systems Engineering from Sapienza University of Rome in 1997. In 2012 and 2005 she has been Visiting Scholar respectively at the Courant Institute, New York University, and at the Robotics Institute, Carnegie Mellon University.

Her research interests are in robot motion planning and control, with emphasis on non-holonomic and redundant systems of which wheeled mobile robots and humanoids are respectively the most notable examples. In particular, over the years her research activity has dealt with visual control, localization, motion planning, physical interaction for humanoids; modeling and control of UAVs; sensor-based exploration of unknown environments with teams of mobile robots; task-constrained motion planning for kinematically redundant systems; probabilistic methods for robot motion planning and exploration; planning and control of wheeled mobile robots; steering and stabilization of general nonholonomic systems; nilpotent approximations of systems with singularities.

On these research themes she has published more than 60 papers in international conferences and journals. The total number of citations received is 2142 according to Google Scholar.

From January 2010 to December 2013 she has served as Associate Editor for the IEEE Transactions on Robotics. From 2009 to 2012 she has been Associate Editor in the ICRA Conference Editorial Board and from 2008 to 2011 for the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). She has contributed to the organization of several international scientific events among which ICRA 2007 as Registration Chair.

She participates to several EC funded projects, among which SAPHARI (FP7), COMANOID (H2020), SIMPLEXITY (H2020).


Proving that Programs do not Discriminate

Oct 10, 2016, 4-5pm, 250 SDH, Aws Albarghouthi, University of Wisconsin-Madison.

Slides

Abstract

Programs have become powerful arbitrators of a range of significant decisions with far-reaching societal impact -- hiring, welfare allocation, prison sentencing, policing, amongst an ever-growing list. In such scenarios, the program is carrying out a sensitive task, and could potentially be illegally discriminating —- advertently or inadvertently —- against a protected group, e.g., African Americans in the United States.

With the range and sensitivity of algorithmic decisions expanding by the day, the question of whether an algorithm is fair (unbiased) has captured the attention of a broad spectrum of experts, from government regulators and law scholars to computer science theorists. Ultimately, algorithmic fairness is a question about programs and their properties: Does a program discriminate against a subset of the population? In this talk, I will view algorithmic fairness through the lens of program verification. Specifically, I will begin by formalizing the notion of fairness as a probabilistic property of programs. To enable automated verification of fairness, I will show how to reduce the probabilistic verification question to that of volume computation over first-order formulas, and describe a new symbolic volume computation algorithm. Finally, I will present results of applying FairSquare -- the first fairness verification tool -- to a variety of decision-making programs.

Bio:

Aws Albarghouthi is an Assistant Professor of Computer Science at the University of Wisconsin-Madison. He works on software verification, analysis, and synthesis.


Oct 17, 2016, 4-5pm, 250 SDH, Tom Griffiths, University of California, Berkeley.

Slides

Abstract

Bio:


Oct 24, 2016, 4-5pm, 250 SDH, Nancy Amato, Texas A&M University.

Slides

Abstract

Bio:


Oct 31, 2016, 4-5pm, 250 SDH, Sergey Levine, University of California, Berkeley.

Slides

Abstract

Bio:


Nov 07, 2016, 4-5pm, 250 SDH, Ian Mitchell, University of British Columbia.

Slides

Abstract

Bio:


Nov 14, 2016, 4-5pm, 250 SDH, Marco Pavone, Stanford University.

Slides

Abstract

Bio:


Nov 21, 2016, 4-5pm, 250 SDH, Hamsa Balakrishnan, Massachusettes Institute of Technology.

Slides

Abstract

Bio:


Nov 28, 2016, 3-4pm, 250 SDH, Karl Johansson, KTH Royal Institute of Technology.

Slides

Abstract

Bio:


Dec 05, 2016, 4-5pm, 250 SDH, Yon Visell, University of California, Santa Barbara.

Slides

Abstract

Bio:


This group has the following subpages:

You are not logged in 
Contact 
©2002-2016 U.C. Regents