Two-player zero-sum differential games with one-sided information and state constraints, aka Football

Date: October 11, 2024; Time: 2:30 PM Location: PWEB 175

Abstract: Enabling embodied intelligence requires robots to plan according to unknown and potentially adversarial intents of interacting agents. This talk will focus on one scenario where theory and methods are underdeveloped. Specifically, we study zero-sum differential games with state constraints and one-sided information, where the informed player (Player 1) has a categorical payoff type unknown to the uninformed player (Player 2). The goal of Player 1 is to minimize his payoff without violating the constraints, while Player 2 either aims to violate the state constraints or, failing that, maximize the payoff. Examples of such games include man-to-man matchup in football and missile defense scenarios. Due to the zero-sum nature, Player 1 may need to delay information release or even manipulate Player 2’s belief to take full advantage of information asymmetry, while Player 2’s strategy will need to balance all possible consequences. Existing solvers such as CFR+ (e.g., for Poker) are applicable, but are not scalable to continuous action spaces as is often the case in robotics. We will discuss efficient solvers for these games by leveraging unique structural properties of their value functions.

Biographical Sketch: Dr. Yi Ren is an Associate Professor in the Department of Mechanical and Aerospace Engineering at Arizona State University. His research spans a range of topics at the intersection of machine learning and engineering, with recent focuses on differential game theory, GenAI model attribution, and representation learning for materials. He has published in both machine learning conferences, including ICLR and ICML, and engineering journals, such as IEEE Transactions on Robotics and Acta Materialia. Dr. Ren received his Ph.D. in Mechanical Engineering from the University of Michigan in 2012 and his Bachelor’s degree in Automotive Engineering from Tsinghua University in 2007. Outside of research, he enjoys playing soccer and spending time with his children.

Abstract

I will first present a portable setup to generate shock waves using the exploding wire technique. Subsequently, I will showcase how droplets of various kinds (liquid metal, water, and polymeric liquids) interact and breakup in the shock wave and associated flow. I will also show the various instabilities that develop prior to breakup that are universal in nature. Lastly, I will showcase some results on shock-droplet flame interactions with analyses on flame extinction and droplet breakup.

Biographical Sketch

Prof. Saptarshi Basu received his PhD in Mechanical Engineering from University of Connecticut in 2007 with Prof. B. M Cetegen before joining University of Central Florida as an Assistant Professor. In 2010, he relocated to India and joined the prestigious Indian Institute of Science in Bangalore where he is currently the Pratt and Whitney Chair Professor in the Department of Mechanical Engineering.

Prof. Basu primarily works on multiphase systems, especially droplets at multiple length and timescales across multiple application domains ranging from surface patterning to combustion. Recently Prof. Basu have done extensive research on transmission of aerosols during COVID and on the efficacy of facemasks. His research marries fundamental aspects of classical fluid mechanics like vortex dynamics and swirling flows and the more interdisciplinary aspects of interfacial transport as in droplets to offer unprecedented insights into multiphase systems.

He is a fellow of Indian National Academy of Engineering, ASME, Institute of Physics, Royal Aeronautical Society and Royal Society of Chemistry. Prof. Basu is the recipient of DST Swarnajayanti Fellowship (equivalent of PECASE) in Engineering. Prof. Basu is a co-founder of a Biotech startup specializing in AI based Point of Care Diagnostics and a technical advisor to a deep tech startup involved in micro gas turbines. Prof. Basu serves as an editor/guest editor of several journals like Nature-scientific reports, Experiments in Fluids and European Physical Journal Special Topics. Prof. Basu’s research is extensively funded by Department of Defence, Indian Space Research Organization, Department of Science and Technology, Indo-German Science and Technology Center, Indo-US Clean Energy Center, NSF and industries like Siemens and Tata Motors. Prof. Basu has guided more than 20 PhD students in his career and published over 200 journal articles including many in Journal of Fluid Mechanics, Physics of Fluids, Combustion & Flame, Langmuir, Proc. Roy. Soc. etc.

Abstract: Advances in electrical energy storage systems are critical for vehicle electrification, renewable energy integration into the electric grid, and electric aviation. Recent years have witnessed an urgent need to accelerate innovation toward realizing improved and safe utilization of high energy and power densities, for example, in lithium-ion and advanced battery chemistries. These are complex, dynamical systems that include coupled processes encompassing electronic, ionic, and solid-state diffusive transport, electrochemical reactions at electrode/electrolyte interfaces, mechanical stress generation, and thermal transport in porous electrodes. This presentation will highlight the importance of the underlying mechanistic interactions at scale in the design of novel paradigms in exemplar energy storage architectures.

Biographical Sketch: Partha P. Mukherjee is a Professor of Mechanical Engineering and a University Faculty Scholar at Purdue University. His prior appointments include Assistant Professor and Morris E. Foster Faculty Fellow of Mechanical Engineering at Texas A&M University (2012-2017), Staff Scientist at Oak Ridge National Laboratory (2009-2011), Director’s Research Fellow at Los Alamos National Laboratory (2008-2009), and Engineer at Fluent India (currently Ansys Inc., 1999-2003). He received his Ph.D. in Mechanical Engineering from Pennsylvania State University in 2007. His awards include Scialog Fellows’ recognition for advanced energy storage, University Faculty Scholar and Faculty Excellence for Early Career Research awards from Purdue University, The Minerals, Metals & Materials Society Young Leaders Award, and invited presentations at the U.S. National Academy of Engineering Frontiers of Engineering symposium and Gordon Research Conference – Batteries, to name a few. His research interests are focused on mesoscale physics and stochastics of transport, chemistry, and materials interactions, including an emphasis on the broad spectrum of energy storage and conversion.

Abstract: The characterization and quantification of the coupling between flow and flexible structures and dominant oscillation modes remain open problems. Environmental science, energy, structural design, and locomotion applications require a comprehensive understanding of these phenomena. Canopy flows, encompassing extensive arrays of rigid or flexible structures, hold significant interest. Ubiquitous in natural environments and spanning multiple scales, they are instrumental in the transport of scalar and inertial particles. This presentation will provide insights from both theoretical perspectives and controlled laboratory experiments. I will discuss the role of key parameters modulating the unsteady dynamics of flows, individual structures, and canopies. These parameters comprise flow velocity, turbulence, structural stiffness, aspect ratio, tip effects, layout, and submergence within open channel flows. For this purpose, I will present data from particle image velocimetry (PIV), particle tracking velocimetry (PTV), and force balance analyses, highlighting turbulence, motion patterns, and unsteady loads on selected structures.

Biographical Sketch: Dr. Chamorro is an Associate professor in the Department of Mechanical Science and Engineering at the University of Illinois at Urbana-Champaign and is affiliated with the Departments of Aerospace Engineering, Civil and Environmental Engineering, and Geology. His research interests include turbulence, particle dynamics, boundary layer processes, aerodynamics, turbulence and structure interaction, wind energy, marine and hydrokinetic energies, and the development of advanced flow diagnostics. He has published 135 peer-reviewed articles in leading journals, has participated in over 140 presentations in technical symposia, and serves as scientific chair on Energy, Electrical Eng, Electronics and Mechanics (W&T7) at the Research Foundation Flanders (FWO) in Belgium. Chamorro is Associate Editor of the Journal of Renewable and Sustainable Energy, the Journal Frontiers in Energy Research, and the Journal of Energy Engineering. He leads the Renewable Energy and Turbulent Environment group, which uses a versatile experimental approach that combines state-of-the-art techniques, including 2D/3D particle image velocimetry, computer vision, and 3D particle tracking velocimetry.

Abstract: Deep learning approaches have become crucial tools across numerous engineering domains. However, they face various challenges, as they typically depend on representative data and large training datasets. Conversely, condition monitoring data for complex systems often lacks labels and representativeness, posing significant challenges for purely data-driven approaches. Additionally, deep learning models generally struggle in extrapolation regimes, which are common for assets characterized by long service lives and the frequent emergence of new operating regimes. 

In response to these challenges, the integration of physical laws and principles with deep learning methodologies has shown tremendous promise. This presentation will explore a variety of approaches that combine physics-based concepts with deep learning techniques. One focus will be on how incorporating structural inductive biases into learning architectures, such as through physics-enhanced graph neural networks, can address the aforementioned challenges.

To close the loop and bridge machine learning with physics, the talk will delve into novel approaches of symbolic regression through reinforcement learning to uncover symbolic equations.

Biographical Sketch: Olga Fink has been assistant professor of Intelligent Maintenance and Operations Systems at EPFL since March 2022.  Olga’s research focuses on Hybrid Algorithms Fusing Physics-Based Models and Deep Learning Algorithms, Hybrid Operational Digital Twins, Transfer Learning, Self-Supervised Learning, Deep Reinforcement Learning and Multi-Agent Systems for Intelligent Maintenance and Operations of Infrastructure and Complex Assets.

Before joining EPFL faculty, Olga was assistant professor of intelligent maintenance systems at ETH Zurich from 2018 to 2022, being awarded the prestigious professorship grant of the Swiss National Science Foundation (SNSF). Between 2014 and 2018 she was heading the research group “Smart Maintenance” at the Zurich University of Applied Sciences (ZHAW).

Olga received her Ph.D. degree from ETH Zurich with the thesis on “Failure and Degradation Prediction by Artificial Neural Networks: Applications to Railway Systems”, and Diploma degree in industrial engineering from Hamburg University of Technology. She has gained valuable industrial experience as reliability engineer with Stadler Bussnang AG and as reliability and maintenance expert with Pöyry Switzerland Ltd.

Olga has been a member of the BRIDGE Proof of Concept evaluation panel since 2023. Moreover, Olga is serving as an editorial board member of several prestigious journals, including Mechanical Systems and Signal Processing, Engineering Applications of Artificial Intelligence, Reliability Engineering and System Safety and IEEE Sensors Journal.

In 2018, Olga was honored as one of the “Top 100 Women in Business, Switzerland”. Additionally, in 2019, earned the distinction of being recognized as a young scientist of the World Economic Forum. In 2020 and 2021, she was honored r as a young scientist of the World Laureate Forum. In 2023, she was distinguished as a fellow by the Prognostics and Health Management Society.