Archive of past talks presenting ongoing research across CORAL domains.
Safe and compliant robot–environment interaction is difficult when the robot's dynamic parameters are not precisely known. The talk will begin with the fundamentals of adaptive impedance control and fixed-time control, and then bring the two together to present a fixed-time adaptive impedance scheme that regulates the desired interaction behavior without requiring accurate parameter knowledge. The approach uses a filtering technique that eliminates the need for acceleration measurements, along with a new parameter adaptation law that guarantees fixed-time convergence under mild initial- and interval-excitation conditions, rather than the usual persistence-of-excitation condition, with explicit bounds on the convergence time. The talk will close with validation results and a brief discussion of the practical implications for contact-rich manipulation and physical human–robot interaction.
Chayan Kumar Paul received the B.Tech. degree in Electrical Engineering from the Indian Institute of Engineering Science and Technology (IIEST), Shibpur, India, in 2019. He has submitted his thesis towards Ph.D. degree in control and automation with the Department of Electrical Engineering at the Indian Institute of Technology (IIT) Delhi, New Delhi, India. He is currently a research associate at FOCAS Labs, CPS, IISc. His research interests include the safe adaptive control of robot manipulators, impedance control for human-robot interaction, observer design, and finite/fixed-time control methodologies.
The study of multi-agent systems (MAS) and related control architectures is becoming increasingly popular in applications such as power grids, multi-robot systems, IoT systems, and sensor networks. These systems are large-scale, networked, and consist of multiple interdependent agents. In this talk, the consideration is on distributed control of networked MAS with linear time-invariant dynamics and quadratic performance measures, where each agent has access only to local or output feedback information. As a result, full-state feedback controllers are not implementable. Using a game-theoretic framework, this problem is modeled as a networked differential game. Existing works often assume complete state information, which is not feasible in practice. It's shown that even for low-dimensional games, computing an output feedback Nash equilibrium (OFNE) is challenging, as it involves solving coupled algebraic Riccati equations (CAREs) along with verification of additional structural constraints. To address this, the notion of output feedback guaranteed cost equilibrium (OFGCE) is introduced, where OFGCE controllers achieve an upper bound on individual costs while retaining equilibrium properties. Its key properties, including monotonicity and the Price of Stability, will be discussed. Sufficient conditions for existence and verification, along with an iterative algorithm for computation of an OFGCE will be presented.
The second part of the talk will focus on MAS under external deterministic disturbances. Despite the importance of robustness in control theory, there is limited work on differential games with disturbances. A notion of soft-constrained OFGCE (SC-OFGCE) under quadratically bounded disturbances is introduced. Sufficient conditions and an iterative LMI-based synthesis algorithm will be presented.
Aniruddha Roy is currently a National Postdoctoral Fellow at the Robert Bosch Centre for Cyber-Physical Systems, Indian Institute of Science, Bengaluru, where he works with Prof. Pavan Tallapragada. His current research lies at the intersection
of dynamic games, data-driven game theory, and their applications in networked control systems and target–attacker–defender games. He completed his Ph.D. from Indian Institute of Technology-Madras (IIT M), Chennai, in July 2024 under the supervision of Prof.
Puduru Viswanadha Reddy. His doctoral research developed a novel solution concept, guaranteed cost equilibrium, for a class of linear-quadratic differential games. Before joining IISc, he briefly worked with an aerospace startup and IIT Kharagpur. He received
his M.E. in Control Systems Engineering from Jadavpur University, Kolkata, in 2018, and his B.Tech. in Electrical Engineering from Maulana Abul Kalam Azad University of Technology in 2016. He is a recipient of several prestigious awards and recognitions, including
the ANRF National Postdoctoral Fellowship, the 1980 Batch Project-Excellence in Research Travel Grant from IIT Madras, the Amul Vidya Bhushan Award for outstanding academic excellence among Class XII students across India, and the Charusila Memorial Award
from Midnapore Town School, West Bengal, for achieving the highest marks in Class XII.
The embedded closed-loop structure of integrated systems can be compromised when
adversaries successfully launch malicious attacks. Consequently, over the past decade,
significant attention has been devoted to enhancing the safety and security of CyberPhysical Systems (CPSs). This article investigates the resilient control problem for CPSs
with multiple transmission channels, where both sensor-to-controller and controller-toactuator links are subject to Denial-of-Service (DoS) attacks.
First, a switched observer is developed to address the Multi-Channel DoS (MCDoS)
scenario. Next, the admissible changing frequency of MCDoS is characterized while
ensuring Input-to-State Stability (ISS) of the closed-loop system. Subsequently, an eventtriggered control strategy is proposed to mitigate the effects of Full-Scale DoS (FSDoS).
Finally, the control parameters are optimally designed to maximize the system’s
resilience against the highest allowable MCDoS changing frequency.
Anindya Basu is currently an Institute Postdoctoral Fellow in the Department of CyberPhysical Systems at the Indian Institute of Science (IISc), Bangalore. He received his
Ph.D. in 2026 from the Department of Electronics and Electrical Engineering at IIT
Guwahati, where he was affiliated with the Systems, Control, and Automation group. His
research focuses on the safety and security of cyber-physical systems, with particular
emphasis on resilient control design for networked control systems under denial-ofservice (DoS) attacks, event-triggered control, and model predictive control
Mirror descent extends gradient-based optimization to spaces equipped with a non-Euclidean geometry, typically induced by a Bregman divergence. In this talk, we revisit this framework through the lens of projected dynamical systems and highlight its implications in stochastic first- and zeroth-order settings.
We begin with a continuous-time viewpoint. In particular, we present a projected dynamical system defined on a Riemannian manifold and show that mirror descent arises as its Euler discretization. This connection provides a natural geometric interpretation of the algorithm.
We then move to the stochastic setting, where only noisy gradient information is available, possibly generated by an underlying Markov process. A key insight is that the almost sure convergence of stochastic mirror descent can be understood through the asymptotic stability of the underlying deterministic projected dynamical system. This viewpoint enables analysis using tools from dynamical systems theory.
Next, we consider the zeroth-order setting, where only noisy function evaluations are accessible. By constructing suitable gradient approximations, we show that the resulting stochastic zeroth-order mirror descent can be interpreted as a perturbed projected dynamical system. Using ideas from robust stability analysis, we establish its almost sure convergence.
Overall, the talk develops a unified perspective connecting mirror descent, projected dynamical systems in non-Euclidean domains, and stochastic approximation. We conclude with finite-time guarantees and briefly discuss connections to reinforcement learning, distributionally robust optimization, and nonsmooth nonconvex problems.
Anik Kumar Paul is currently a Walmart Postdoctoral Fellow in the Department of Computer Science and Automation at the Indian Institute of Science (IISc), Bangalore. He received his Ph.D. in 2024 from the Department of Electrical Engineering at IIT Madras, where he was part of the Control and Optimization group. His research interests include stochastic optimization, distributed learning, and nonsmooth analysis.
Bearing-only formation control has garnered immense research attraction because bearing measurements can be obtained using low-cost cameras, require less complex sensory arrangement, are less noisy, and are desirable for military applications. While specifying a desired formation with bearing constraints, the agents are required to sense their orientations for the implementation. However, elevation angle constraints, which can be obtained using vision-only sensors, are independent of coordinate frames. Hence, formation control using elevation angle constraints does not require information about a global frame of reference, or any orientation synchronization or orientation estimation protocol. Recently proposed gradient-based bearing-only control laws using elevation angle constraints have only established local stability results with single integrator dynamics. In this talk, robust control laws for single integrators and double integrators will be discussed when the agents' dynamics are affected by unknown bounded disturbances. One problem with using only elevation angle constraints to define the desired formation is that the stability results are only local. Hence, the agents may converge to undesired equilibria if the initial conditions are not close to the desired formation shape. To mitigate this issue, a new rigidity theory, named “sign-elevation angle rigidity,” will be developed that uses signed-area and volume constraints along with elevation angle constraints to describe the desired formation. Then, how to obtain an almost global stability result for the elevation angle-based approach will be discussed.
Chinmay Garanayak obtained his PhD from the Department of Electrical Engineering, Indian Institute of Technology Bombay, in 2024. He received his B.Tech. (2015) from the National Institute of Technology, Rourkela, India, in Electrical Engineering. He was a postdoctoral researcher at the Department of Electrical Engineering, IIT Bombay till 2025. Currently, he is a postdoctoral researcher at the Department of Electrical Engineering, NYU Abu Dhabi. His research interests include multi-agent systems, cooperative control, formation control, and nonlinear control.
In this interactive session, we will embark on a captivating exploration of the remarkable world of Internet of Things (IoT) systems that have some extreme characteristics. These systems hold immense potential to benefit us even in applications that confront extreme limits. While the IoT has already revolutionised our interaction with technology by connecting everyday objects to the digital world, this talk takes us beyond conventional boundaries. We will dive deeper into the IoT systems, where we push the limits and discover how IoT can thrive and excel in environments with traditional constraints. Through a collage of multiple demonstrable IoT systems, we will explore various aspects, including crucial factors such as battery life, longevity, delay, and the challenging environmental conditions in which these systems must operate. By challenging the status quo, we uncover novel solutions that overcome these hurdles and unleash the true potential of sensors and radios. Prepare to be inspired as we present real-world examples and showcase research that demonstrates the transformative power of this field. Through this interactive session, we will, together, engage, ponder, and envision the future possibilities that lie within the realm of IoT Systems.
Dr. R Venkatesha Prasad is an associate professor at the Networked Systems group of Delft University of Technology (TU Delft). At TU Delft, he has supervised 23 PhD students and 90 MSc students. He has (co-)authored more than 350 publications in peer-reviewed international transactions/journals, Patents, and conferences in the areas of Tactile Internet, Internet of Things (IoT), Cyber Physical Systems (CPS), Energy-harvesting, 60 GHz mmWave networks, Smart energy systems, Personal networks, Cognitive Radios and Voice over Internet Protocol (VoIP). Recognising his research contributions to IoT, he has been selected as an IEEE Communications Society (ComSoc) Distinguished Lecturer on the Internet of Things for the period 2016-2020. He is currently mentoring IEEE Tactile Internet and IEEE 6G Robotics standardisation groups. He is also leading many IEEE activities through positions on standards boards and technical committees. He is on the editorial board of many IEEE and international transactions and magazines. He was the Deputy Project Director for Lunar Zebro – a moon rover project. He completed his PhD from IISc, Bangalore, India, in 2003. His thesis work led to a start-up venture, Esqube Communication Solutions, which was selected as one of the top 100 IT innovators in India in 2006 by NASSCOM and the top 100 promising companies in Asia by RedHerring in 2008. He led a student team to win the worldwide Airbus challenge -- Fly Your Ideas. He is a co-founder of ZED BV in the Netherlands and Skynetics India. He is a senior member of IEEE and ACM, and a Fellow of IETE.
Despite remarkable progress in robotics, we still see very few robots actually helping us in our daily lives. A key reason is that real-world deployment imposes fundamental constraints that are often overlooked in controlled demonstrations: robot data is expensive and computation is limited. To meet these challenges we must design resource-rational robots that dynamically adapt their planning and learning to the available resources. I will present a principled framework for robot planning and learning that treats time, data, and computation as resources with explicit costs and budgets that must be allocated intelligently. I will introduce metareasoning algorithms that decide what and how to learn at deployment time, enabling robots in factories and homes to provably maximize their performance under operational constraints. I will then show how this framework enables robots to acquire contact-rich tasks through a few hours of unsupervised practice, and to coordinate dozens of robots in warehouses without any multi-robot data. Together, these results chart a path towards robots that act reliably and learn efficiently under real-world constraints, bringing us closer to long-term autonomy.
Shivam Vats is a postdoctoral researcher in the Department of Computer Science at Brown University, where he works with George Konidaris on developing planning algorithms that continually learn from experience. His research on manipulation, human-robot teaming and multi-robot systems has been recognized with an Outstanding HRI Paper Finalist Award at ICRA 2022 and a Spotlight presentation at ICLR 2025. He earned his PhD in robotics from Carnegie Mellon University, where he was co-advised by Maxim Likhachev and Oliver Kroemer. Shivam also holds a BSc and an MSc in Mathematics and Computing from IIT Kharagpur.
Satisfaction of state and input constraints is a fundamental requirement in many safety-critical control applications, particularly in the presence of uncertainties and disturbances. This talk presents an adaptive control framework that guarantees user-defined constraints on system states and control inputs for uncertain dynamical systems. The proposed approach integrates a barrier Lyapunov function with a saturated control law to ensure constraint satisfaction while maintaining stable tracking, without relying on real-time optimization. An offline verifiable feasibility condition is derived to certify whether a given set of constraints can be safely enforced by the controller. An extension of this framework to handle time-varying state and input constraints in the presence of parametric uncertainties and disturbances will also be briefly discussed. The resulting framework provides a computationally efficient approach for ensuring safety, feasibility, robustness, and performance in uncertain dynamical systems.
Poulomee Ghosh is currently a Research Associate with Dr. Pushpak Jagtap at the Formal Control and Autonomous Systems (FOCAS) Lab, Robert Bosch Centre for Cyber-Physical Systems (RBCCPS), IISc Bangalore. She recently submitted her Ph.D. thesis in Electrical Engineering at the Indian Institute of Technology Delhi under the supervision of Prof. Shubhendu Bhasin. Her doctoral research focuses on the development of adaptive control strategies for uncertain dynamical systems, with guarantees on user-defined constraints on system states, control inputs, and input rates. She received her B.Tech. in Electrical Engineering from the National Institute of Technology Durgapur in 2020.