Safe Control of Large-scale MAS: Recent Developments and Open Challenges

Room: Churchill B1

Overview

Multi-agent systems (MAS) have received significant attention from scholars across disciplines, including computer science and robotics, as a means of solving complex problems by subdividing them into smaller tasks. Some examples of MAS include smart grids, search-and-rescue teams, edge computing, wireless communication networks, space systems, package delivery, and microgrids. The design and analysis of MAS controllers present unique challenges, including scalability, verification, and robustness against factors such as communication issues, adversarial or non-cooperative agents, and partial observability. The most common tools for the design and analysis of safety in MAS include shielding-based methods, safety learning certificates, such as control barrier functions, and reinforcement learning-based methods. When it comes to the performance of these tools, there is generally a trade-off between scalability and the ability to formally verify their correctness. While there has been significant progress in the field toward mitigating this tradeoff, much work remains to develop a formally verifiable method that is both scalable and practically useful. The main goal of the workshop is to highlight recent advances and developments in the role of control theory and machine learning in addressing scalability issues in safe control design for MAS, and to discuss key open problems in the field. This workshop aims to bring together experts from machine learning, robotics, and control theory to discuss recent developments and the utility of available tools for safe MAS, along with their drawbacks and directions for future research, given the current state of the art. 

Workshop schedule

• 9:00 am to 9:10 am - Opening remarks

• 9:10 pm to 10:am pm - Kunal Garg - "Verification of scalable neural CBFs for safe multi-agent robot systems"

• 10:00 am to 10:30 am - Coffee break

• 10:30 am to 11:15 am - Jonathan How - "Safe Multi-Agent Planning and Control Under Uncertainty"

• 11:15 am to 12:00 pm - Lars Lindemann - "When Control Changes the Data: Safety under Interaction-Driven Distribution Shifts"

• 12:00 pm to 2:00 pm - Lunch break

• 2:00 pm to 3:00 pm - Dimitra Panagou - "Multi-Agent Autonomy for Resilient Embodied AI"

• 3:00 pm to 3:30 pm - Coffee break

• 3:30 pm to 4:20 pm - Samuel Coogan - "Interval MDPs for Safe, Learning-Aware Multiagent Systems"

• 4:20 pm to 5:00 pm - Xi Yu - "From Connectivity to 'Maintainability': When the Environment Challenges Multi-Robot Systems"

• 5:00 pm to 5:45 pm - Ruiting Wang - "Sociotechnical Infrastructures Design for the Energy-Mobility Nexus"

Talk info:

Kunal Garg: "Verification of scalable neural CBFs for safe multi-agent robot systems"

Abstract: Designing controllers with provable safety guarantees and scalability for large-scale multi-agent systems (MAS) is a challenging problem. A variety of methods have been proposed in recent years, utilizing the success of machine learning (ML)-based approaches, such as certificate or safety filter learning methods. In this tutorial-style talk, we will cover the methods based on learning control barrier functions (CBFs) and their extensions to multi-agent systems, such as graph CBF (GCBF). We will talk about the theoretical guarantees provided by these methods and the advances in the training frameworks used specifically for learning CBFs, followed by a summary of their experimental results. Lastly, we will talk about their limitations and the open challenges in being able to provide provable guarantees for the safety of large-scale MAS, particularly when using ML components in the control architecture, with recent developments on probing-based methods as a step towards explainability (if not formal verification) of neural networks in safety-critical applications. 

Bio: Kunal Garg is an assistant professor in the mechanical and aerospace engineering program at the School for Engineering of Matter, Transport and Energy at ASU. He received his master's degree in engineering and doctoral degree in aerospace engineering from the University of Michigan in 2019 and 2021. Before joining ASU, he was a postdoctoral associate in the Laboratory for Information & Decision Systems (LIDS) and the Department of Aeronautics and Astronautics at the Massachusetts Institute of Technology. Garg was a 2022 DAAD AInet Fellow for the Postdoctoral Networking Tour in Artificial Intelligence in the AI and Robotics domain. He also received the Professor Pierre T. Kabamba Award for Excellence in Control Systems, and the Richard and Eleanor Towner Prize for Distinguished Academic Achievement for his PhD research work in 2021. Garg's research interests include robust control synthesis for multi-agent coordination using machine learning methods, finite- and fixed-time control synthesis for spatiotemporal specifications, and continuous-time optimization.

Jonathan How: "Safe Multi-Agent Planning and Control Under Uncertainty"

Bio: Jonathan P. How is the Richard C. Maclaurin Professor of Aeronautics and Astronautics at the Massachusetts Institute of Technology. He received a B.A.Sc. (aerospace) from the University of Toronto in 1987, and his S.M. and Ph.D. in Aeronautics and Astronautics from MIT in 1990 and 1993, respectively, and then studied for 1.5 years at MIT as a postdoctoral associate. Prior to joining MIT in 2000, he was an assistant professor in the Department of Aeronautics and Astronautics at Stanford University. He is a Fellow of the Institute of Electrical and Electronics Engineers (IEEE) and the American Institute of Aeronautics and Astronautics (AIAA). He was elected to the National Academy of Engineering (NAE) in 2021. Dr. How’s research focuses on robust planning and learning under uncertainty with an emphasis on multi-agent systems, and he was the planning and control lead for the MIT DARPA Urban Challenge team in 2007. His work has been recognized with multiple awards, including the 2020 IEEE CSS Distinguished Member Award, the 2020 AIAA Intelligent Systems Award, the 2002 Institute of Navigation Burka Award, the 2011 IFAC Automatica award for best applications paper, the 2015 AeroLion Technologies Outstanding Paper Award for Unmanned Systems, the 2015 winner of the IEEE Control Systems Society Video Clip Contest, the IROS Best Paper Award on Cognitive Robotics (2017 and 2019), the 2020 ICRA Best Paper Award in Service Robotics, and three AIAA Best Paper in Conference Awards (2011-2013). He received the Amazon Machine Learning Research Award in 2018 and 2020, and he was awarded the Air Force Commander’s Public Service Award in 2017 for his contributions to the SAB. Dr. How serves on the AIAA Fellows Committee for Information Systems and Systems Integration and is the 2026-2028 president-elect for the IEEE Control System Society.

Lars Lindemann: "When Control Changes the Data: Safety under Interaction-Driven Distribution Shifts"

Abstract: Accelerated by rapid advances in machine learning and AI, there has been tremendous success in the design of learning-enabled autonomous systems in areas such as autonomous driving and robotics. These exciting developments are accompanied by new fundamental challenges that arise regarding the safety and reliability of these increasingly complex systems due to imperfect learning, system unknowns, and uncertain environments. Statistical tools for uncertainty quantification have gained popularity due to their ability to deal with these challenges. However, their guarantees rely on i.i.d. data, an assumption that is violated when control actions change the underlying data distribution. In this talk, I will provide new insight to design safe controllers under distribution shifts using robust conformal prediction (CP). I will begin by advocating for the use of CP due to its simplicity, generality, and efficiency as opposed to existing optimization-based verification techniques. I will then provide an introduction to CP and summarize existing work that uses CP to design probabilistically safe controllers in dynamic environments. Subsequently, we will look into interactive settings where the system’s behavior may change the environment's behavior, and vice versa. This circular dependency creates an interaction-driven distribution shift that invalidates existing CP guarantees. To deal with this problem, we propose an iterative framework that episodically updates the controller while robustly maintaining safety guarantees by quantifying the potential impact of a controller update on the environment's behavior. We realize this via adversarially robust CP where we perform a regular CP step in each episode using observed data under the current controller, but then transfer safety guarantees across controller updates by analytically adjusting the CP result to account for distribution shifts. Lastly, I will show how these ideas extend to handling policy-induced distribution shifts that arise when using barrier/Lyapunov functions to control uncertain systems.

Bio: Lars is an Assistant Professor at the Automatic Control Laboratory at ETH Zürich where he is leading the Algorithmic Systems Theory (AST) Laboratory. His research interests include systems and control theory, formal methods, and machine learning with applications in autonomous systems and robotics. From 2023 to 2025 he was an Assistant Professor in the Thomas Lord Department of Computer Science at the University of Southern California. Between 2020 and 2022, he was a Postdoctoral Fellow in the Department of Electrical and Systems Engineering at the University of Pennsylvania. He received the Ph.D. degree in Electrical Engineering from KTH Royal Institute of Technology in 2020. Prior to that, he received the M.Sc. Degree in Systems, Control and Robotics from KTH in 2020 and two B.Sc. Degrees in Electrical and Information Engineering and in Engineering Management from the Christian-Albrecht University of Kiel in 2014. He received the Outstanding Student Paper Award at the 58th IEEE Conference on Decision and Control and the Student Best Paper Award (as an advisor) at the 60th IEEE Conference on Decision and Control. He was a finalist for the Best Paper Award (as an advisor) at the 2024 International Conference on Cyber-Physical Systems, the Best Paper Award at the 2022 Conference on Hybrid Systems: Computation and Control, and the Best Student Paper Award at the 2018 American Control Conference.

Dimitra Panagou: "Multi-Agent Autonomy for Resilient Embodied AI"

Abstract: Resilience against failures, errors and attacks is a fundamental property that allows a system to “keep running” even at reduced functionality, and is particularly timely for systems that are driven or enabled by AI (embodied or physical AI). Research on resilience and related concepts such as fault tolerance involve distributed learning, planning and control approaches and has been vivid across multiple communities including robotics and control. Multi-robot systems in particular pose significant challenges given that attacks or failures can occur either at the "cyber" domain, e.g., the information shared via communication among robots or acquired via onboard sensing, or at the "physical" domain, e.g., at the vital components of each robot (sensors, actuators) or to the entire system as a whole. Despite tremendous progress, there are still open problems, including but not limited to how we can ensure the safe, secure and efficient operation of the robots in challenging environments, e.g., subject to internal and external constraints. In this talk, I will present an overview and highlights of our recent work on resilient multi-robot systems with a focus on safety, security and efficiency, as well as connections with new avenues such as how to enable resilient AI systems. 

Bio: Dimitra Panagou received the Diploma and PhD degrees in Mechanical Engineering from the National Technical University of Athens, Greece, in 2006 and 2012, respectively. In September 2014 she joined the Department of Aerospace Engineering, University of Michigan as an Assistant Professor. Since July 2022 she is an Associate Professor with the newly established Department of Robotics, with a courtesy appointment with the Department of Aerospace Engineering, University of Michigan. Prior to joining the University of Michigan, she was a postdoctoral research associate with the Coordinated Science Laboratory, University of Illinois, Urbana-Champaign (2012-2014), a visiting research scholar with the GRASP Lab, University of Pennsylvania (June 2013, Fall 2010) and a visiting research scholar with the University of Delaware, Mechanical Engineering Department (Spring 2009). Her research program spans the areas of nonlinear systems and control; multi-agent systems, autonomy and control; and aerospace robotics. She is particularly interested in the development of provably-correct methods for the safe and secure (resilient) operation of autonomous systems in complex missions, with applications in robot/sensor networks and multi-vehicle systems (ground, marine, aerial, space) under uncertainty. She is a recipient of the NASA Early Career Faculty Award, the AFOSR Young Investigator Award, the NSF CAREER Award, the George J. Huebner, Jr. Research Excellence Award, and a Senior Member of the IEEE and the AIAA.

Samuel Coogan: "Interval MDPs for Safe, Learning-Aware Multiagent Systems"

Abstract: Deploying robot teams in real environments requires planning under uncertainty: terrain, dynamics, and obstacle locations are rarely fully known in advance, yet safety constraints must still be enforced. This talk presents interval Markov decision processes (IMDPs) as a unifying tool for safe, learning-aware multiagent planning. We first show how partially known continuous dynamics can be abstracted into an IMDP using Gaussian process regression, yielding certified transition-probability intervals that provably contain the true system behavior. We then present an algorithm for synthesizing controllers that maximize the probability of satisfying rich, LTL-specified objectives on the resulting IMDP. In the second half of the talk we scale this framework to heterogeneous bipedal robot teams operating on uncertain terrain. A hierarchical decomposition assigns each robot its own IMDP planner while a task-level coordinator handles precedence-constrained multi-robot missions. Validated in MuJoCo, three-robot teams complete a ten-task package relay in less than half the single-robot time.

Bio: Sam Coogan received the B.S. degree in Electrical Engineering from Georgia Tech and the M.S. and Ph.D. degrees in Electrical Engineering from the University of California, Berkeley. In 2015, he was a postdoctoral research engineer at Sensys Networks, Inc., and in 2012 he spent time at NASA's Jet Propulsion Lab. Before joining Georgia Tech in 2017, he was an assistant professor in the Electrical Engineering department at UCLA from 2015–2017. His awards and recognitions include the 2020 Donald P Eckman Award from the American Automatic Control Council recognizing "an outstanding young engineer in the field of automatic control", a Young Investigator Award from the Air Force Office of Scientific Research in 2019, a CAREER Award from the National Science Foundation in 2018, and the Outstanding paper award for the IEEE Transactions on Control of Network Systems in 2017.

Xi Yu: "From Connectivity to 'Maintainability': When the Environment Challenges Multi-Robot Systems"

Abstract: Large-scale multi-robot systems are most valuable in complex, uncertain, or hazardous environments where a single robot is insufficient. Yet these are precisely the environments that make connectivity difficult to maintain: sensing is local and noisy, communication is unreliable, actuation is limited, and robot motion is affected by drift and disturbances. This talk reexamines connectivity maintenance through the lens of safety control and physical feasibility. Rather than asking only whether a team is connected in an abstract graph, we ask which connections are actually supportable by robots operating under environmental and hardware constraints. This shift from connectivity to maintainability leads to formulations that explicitly account for sensing limits, communication constraints, actuation capabilities, and environmental dynamics. The resulting perspective supports scalable multi-robot coordination that is not only connected in theory, but maintainable in practice.

Bio: Xi Yu is an assistant professor in the School of Manufacturing Systems and Networks. Xi received her MS and PhD degrees from the Department of Mechanical Engineering at Boston University. She previously worked as a postdoc researcher at the GRASP Lab at the University of Pennsylvania and as an assistant professor in mechanical and aerospace engineering at West Virginia University. With expertise in multi-robot systems, autonomy in extreme environments, and autonomous systems, Yu is passionate about bridging the gap between engineering concepts and real-world applications. Her research interests include multi-agent systems, resilient autonomous systems, and optimal control and decision-making.

Ruiting Wang: "Sociotechnical Infrastructures Design for the Energy-Mobility Nexus"

Abstract: The energy evolutions in mobility are unfolding faster than the power grids and transportation networks meant to support them. This presentation outlines systemic strategies to address infrastructure issues in the electrification of different mobility systems, ranging from the planning challenges in logistics systems to the societal challenges in urban mobility. Specifically, we look at infrastructure network design by examining the underlying decision-making process. Finally, we discuss the fundamentally decentralized nature of these problems and the resulting scalability issues.

Bio: Ruiting Wang is currently a Postdoctoral Fellow at KTH Royal Institute of Technology. She received her Ph.D. degree in Systems Engineering from the University of California, Berkeley, and her B.S. degree in Building Energy Engineering from Tsinghua University. Her work bridges technology and policy, using control, optimization, and machine learning to design robust societal-scale systems with applications in urban energy systems, intelligent transportation networks, and electric vehicle markets. She was selected as a Rising Star in Mechanical Engineering by MIT, and a Trailblazer in Engineering Fellow by Purdue University in 2025. She also the recipient of the Energy System Best Paper award (as student) at American Control Conference 2024, and Best Paper Award (as mentor) at IEEE Forum for Innovative Sustainable Transportation Systems 2026.

Organizers

Kunal Garg, School for Engineering of Matter, Transport, and Energy, ASU (kgarg24@asu.edu)

Chuchu Fan, Department of Aeronautics and Astronautics, MIT (chuchu@mit.edu)