About

Claire Tomlin is a Professor of Electrical Engineering and Computer Sciences at UC Berkeley, holding the Charles A. Desoer Chair in the College of Engineering. Her research interests include hybrid systems, distributed and decentralized optimization, and control theory, with a focus on applications such as unmanned aerial vehicles, air traffic control, and modeling of biological processes. She has contributed significantly to the understanding and development of control systems, particularly in the context of safety verification, collision avoidance protocols, and cyber-physical systems. Her academic career includes teaching at Stanford University from 1998 to 2007, where she was a director of the Hybrid Systems Laboratory and held joint positions in the Department of Aeronautics and Astronautics and the Department of Electrical Engineering. She has been recognized with numerous awards, including a MacArthur Genius grant in 2006 and the IEEE Transportation Technologies Award in 2017 for her contributions to air transportation systems. Dr. Tomlin is a member of the American Academy of Arts and Sciences and the National Academy of Engineering, and she is an IEEE Fellow, reflecting her distinguished impact in her field.

Research topics

• Computer Science
• Artificial Intelligence
• Mathematics
• Algorithm
• Mathematical optimization
• Engineering
• Political Science
• Machine Learning
• Medical education
• Physics
• Geometry
• Internal medicine
• Engineering physics
• Oncology
• Engineering ethics
• Control engineering
• Medicine
• Mathematical analysis

Selected publications

• Active Calibration of Reachable Sets Using Approximate Pick-to-Learn

  ArXiv.org · 2026-03-26

  articleOpen accessSenior author

  Reachability computations that rely on learned or estimated models require calibration in order to uphold confidence about their guarantees. Calibration generally involves sampling scenarios inside the reachable set. However, producing reasonable probabilistic guarantees may require many samples, which can be costly. To remedy this, we propose that calibration of reachable sets be performed using active learning strategies. In order to produce a probabilistic guarantee on the active learning, we adapt the Pick-to-Learn algorithm, which produces generalization bounds for standard supervised learning, to the active learning setting. Our method, Approximate Pick-to-Learn, treats the process of choosing data samples as maximizing an approximate error function. We can then use conformal prediction to ensure that the approximate error is close to the true model error. We demonstrate our technique for a simulated drone racing example in which learning is used to provide an initial guess of the reachable tube. Our method requires fewer samples to calibrate the model and provides more accurate sets than the baselines. We simultaneously provide tight generalization bounds.

  PublisherOA PDF

• Active Calibration of Reachable Sets Using Approximate Pick-to-Learn

  arXiv (Cornell University) · 2026-03-26

  preprintOpen accessSenior author

  Reachability computations that rely on learned or estimated models require calibration in order to uphold confidence about their guarantees. Calibration generally involves sampling scenarios inside the reachable set. However, producing reasonable probabilistic guarantees may require many samples, which can be costly. To remedy this, we propose that calibration of reachable sets be performed using active learning strategies. In order to produce a probabilistic guarantee on the active learning, we adapt the Pick-to-Learn algorithm, which produces generalization bounds for standard supervised learning, to the active learning setting. Our method, Approximate Pick-to-Learn, treats the process of choosing data samples as maximizing an approximate error function. We can then use conformal prediction to ensure that the approximate error is close to the true model error. We demonstrate our technique for a simulated drone racing example in which learning is used to provide an initial guess of the reachable tube. Our method requires fewer samples to calibrate the model and provides more accurate sets than the baselines. We simultaneously provide tight generalization bounds.

  PublisherDOI

• Recursively Feasible Probabilistic Safe Online Learning With Control Barrier Functions

  IEEE Open Journal of Control Systems · 2025-01-01

  articleOpen access

  Learning-based control has demonstrated great promise for handling complex tasks in various applications. However, ensuring system safety under uncertain dynamics remains a significant challenge. Control Barrier Functions (CBFs) offer mathematical tools for enforcing safety constraints given known system dynamics, yet their guarantees can be lost in the presence of model errors. In this article, we present a framework that combines model-based safety methods with data-driven techniques to guarantee safety recursively for systems with uncertain dynamics. We build upon our previous work, where Gaussian Process (GP) regression was utilized to quantify uncertainty in model-based CBF constraints, resulting in a second-order cone program (SOCP) controller. When the SOCP is feasible at a state, it provides a pointwise probabilistic safety guarantee. A critical innovation we develop further in this work is an eventtriggered online data collection algorithm that actively and safely gathers data to provide the recursive feasibility of the SOCP-based controller. By continuously assessing the sufficiency of data based on the feasibility measure of the SOCP, our method triggers safe exploratory actions when necessary to reduce the uncertainty in critical control directions. This approach ensures that a feasible, safety-preserving control input is always available, thereby establishing forward invariance of the safe set with high probability, even in previously unexplored regions. We validate the proposed framework through two numerical simulation experiments.

  PublisherDOI

• Constraint-Guided Online Data Selection for Scalable Data-Driven Safety Filters in Uncertain Robotic Systems

  IEEE Transactions on Robotics · 2025-01-01 · 2 citations

  article

  As the use of autonomous robots expands in tasks that are complex and challenging to model, the demand for robust data-driven control methods that can certify safety and stability in uncertain conditions is increasing. However, the practical implementation of these methods often faces scalability issues due to the growing amount of data points with system complexity, and a significant reliance on high-quality training data. In response to these challenges, this study presents a scalable data-driven controller that efficiently identifies and infers from the most informative data points for implementing data-driven safety filters. Our approach is grounded in the integration of a model-based certificate function-based method and Gaussian Process (GP) regression, reinforced by a novel online data selection algorithm that reduces time complexity from quadratic to linear relative to dataset size. Empirical evidence, gathered from successful real-world cart-pole swing-up experiments and simulated locomotion of a five-link bipedal robot, demonstrates the efficacy of our approach. Our findings reveal that our efficient online data selection algorithm, which strategically selects key data points, enhances the practicality and efficiency of data-driven certifying filters in complex robotic systems, significantly mitigating scalability concerns inherent in nonparametric learning-based control methods.

  PublisherDOI

• Social Planning With the Replicator Dynamics

  IEEE Control Systems Letters · 2025-01-01

  articleSenior author

  Approaches to social planning tend to assume that the behaviour of agents is at an equilibrium, yet in practice people’s behaviour gradually adapts to their experiences. In this work, a model of social planning under the replicator dynamics is studied. This model allows for a social planner to control the learning process of agents by influencing the relative fitness of different strategies. The desiderata that such a social planner would ideally achieve – exponential stability and budget-balance – are described. Existence of a solution for any full-support distribution, as well as an analysis of its properties, are shown constructively by leveraging classical tools from geometric control theory. Though the solution is optimal in an environment without transfer costs, this may not generally hold otherwise. We formulate a relevant optimal control problem to model this setting, and determine performance guarantees based in our original solution.

  PublisherDOI

• Mechanistic interpretability for steering vision-language-action models

  ArXiv.org · 2025-08-30

  preprintOpen accessSenior author

  Vision-Language-Action (VLA) models are a promising path to realizing generalist embodied agents that can quickly adapt to new tasks, modalities, and environments. However, methods for interpreting and steering VLAs fall far short of classical robotics pipelines, which are grounded in explicit models of kinematics, dynamics, and control. This lack of mechanistic insight is a central challenge for deploying learned policies in real-world robotics, where robustness and explainability are critical. Motivated by advances in mechanistic interpretability for large language models, we introduce the first framework for interpreting and steering VLAs via their internal representations, enabling direct intervention in model behavior at inference time. We project feedforward activations within transformer layers onto the token embedding basis, identifying sparse semantic directions - such as speed and direction - that are causally linked to action selection. Leveraging these findings, we introduce a general-purpose activation steering method that modulates behavior in real time, without fine-tuning, reward signals, or environment interaction. We evaluate this method on two recent open-source VLAs, Pi0 and OpenVLA, and demonstrate zero-shot behavioral control in simulation (LIBERO) and on a physical robot (UR5). This work demonstrates that interpretable components of embodied VLAs can be systematically harnessed for control - establishing a new paradigm for transparent and steerable foundation models in robotics.

  PublisherOA PDFDOI

• Resolving Conflicting Constraints in Multi-Agent Reinforcement Learning with Layered Safety

  2025-06-21 · 2 citations

  preprintOpen accessSenior author

  Preventing collisions in multi-robot navigation is crucial for deployment. This requirement hinders the use of learning-based approaches, such as multi-agent reinforcement learning (MARL), on their own due to their lack of safety guarantees. Traditional control methods, such as reachability and control barrier functions, can provide rigorous safety guarantees when interactions are limited only to a small number of robots. However, conflicts between the constraints faced by different agents pose a challenge to safe multi-agent coordination. To overcome this challenge, we propose a method that integrates multiple layers of safety by combining MARL with safety filters. First, MARL is used to learn strategies that minimize multiple agent interactions, where multiple indicates more than two. Particularly, we focus on interactions likely to result in conflicting constraints within the engagement distance. Next, for agents that enter the engagement distance, we prioritize pairs requiring the most urgent corrective actions. Finally, a dedicated safety filter provides tactical corrective actions to resolve these conflicts. Crucially, the design decisions for all layers of this framework are grounded in reachability analysis and a control barrier-value function-based filtering mechanism. We validate our Layered Safe MARL framework in 1) hardware experiments using Crazyflie drones and 2) high-density advanced aerial mobility (AAM) operation scenarios, where agents navigate to designated waypoints while avoiding collisions. The results show that our method significantly reduces conflict while maintaining safety without sacrificing much efficiency (i.e., shorter travel time and distance) compared to baselines that do not incorporate layered safety. The project website is available at https://dinamo-mit.github.io/Layered-Safe-MARL/

  PublisherDOI

• Unfamiliar Finetuning Examples Control How Language Models Hallucinate

  2025-01-01 · 6 citations

  articleOpen access

  Katie Kang, Eric Wallace, Claire Tomlin, Aviral Kumar, Sergey Levine. Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers). 2025.

  PublisherOA PDFDOI

• University of California, Berkeley, Berkeley, CA, USA [Institutes in Control]

  IEEE Control Systems · 2025-09-25

  article1st authorCorresponding
  PublisherDOI

• Explaining Low Perception Model Competency with High-Competency Counterfactuals

  Communications in computer and information science · 2025-10-18

  book-chapterOpen accessSenior author

  Abstract There exist many methods to explain how an image classification model generates its decision, but very little work has explored methods to explain why a classifier might lack confidence in its prediction. As there are various reasons the classifier might lose confidence, it would be valuable for this model to not only indicate its level of uncertainty but also explain why it is uncertain. Counterfactual images have been used to visualize changes that could be made to an image to generate a different classification decision. In this work, we explore the use of counterfactuals to offer an explanation for low model competency–a generalized form of predictive uncertainty that measures confidence. Toward this end, we develop five novel methods to generate high-competency counterfactual images, namely Image Gradient Descent (IGD), Feature Gradient Descent (FGD), Autoencoder Reconstruction (Reco), Latent Gradient Descent (LGD), and Latent Nearest Neighbors (LNN). We evaluate these methods across two unique datasets containing images with six known causes for low model competency and find Reco, LGD, and LNN to be the most promising methods for counterfactual generation. We further evaluate how these three methods can be utilized by pre-trained Multimodal Large Language Models (MLLMs) to generate language explanations for low model competency. We find that the inclusion of a counterfactual image in the language model query greatly increases the ability of the model to generate an accurate explanation for the cause of low model competency, thus demonstrating the utility of counterfactual images in explaining low perception model competency (The code for reproducing our methods and results is available on GitHub: https://github.com/sarapohland/competency-counterfactuals .).

  PublisherOA PDFDOI

Recent grants

• NIH Grant R01GM075311

  NIH · $1.2M · 2010

• CPS: Large: ActionWebs

  NSF · $5.0M · 2009–2016

• Understanding the Impact of Microscale and Nanoscale Heterogeneity and Resistance

  NIH · $20.4M · 2020–2023

• Comparative analysis of PCP signaling architecture

  NIH · $2.8M · 2011–2020

Frequent coauthors

• Somil Bansal

  73 shared

• David Fridovich-Keil

  69 shared

• Mo Chen

  Simon Fraser University

  58 shared

• Anil Aswani

  48 shared

• Jaime F. Fisac

  46 shared

• Roel Dobbe

  41 shared

• S. Shankar Sastry

  39 shared

• Young Hwan Chang

  Oregon Health & Science University

  37 shared

Education

• Ph.D., Electrical Engineering

  Stanford University

  2007

• M.S., Electrical Engineering

  Stanford University

• B.S., Electrical Engineering and Computer Science

  Massachusetts Institute of Technology (MIT)

Awards & honors

• MacArthur Genius grant (2006)
• IEEE Transportation Technologies Award (2017)
• American Academy of Arts and Sciences Member (2019)
• National Academy of Engineering (NAE) Member (2019)
• IEEE Fellow (2010)