Safe Robot Learning


Robots are envisioned to bring improved efficiency and quality to many aspects of our daily lives. Examples include autonomous driving, drone delivery, and service robots. The decision-making in robotic systems often faces various sources of uncertainties (e.g., incomplete sensory information, uncertainties in the environment, interaction with other agents). As data- and learning-based methods continue to gain traction, we must understand how to leverage them in real-world robotic systems in a safe and robust manner to avoid costly hardware failures and/or allow for deployments in the proximity of human operators. Addressing the safety challenges in real-world deployment will require cross-field collaboration. Our goal is to facilitate enduring interdisciplinary partnerships and build a community fostering the advancement of safe robot autonomy research.

Join our mailing list and safe-robot-learning Google group to stay connected!

Review Paper and Simulation Benchmark

Review Paper on Safe Learning in Robotics

Abstract: As data- and learning-based methods gain traction, researchers must also understand how to leverage them in real-world robotic systems, where implementing and guaranteeing safety is imperative—to avoid costly hardware failures and allow deployment in the proximity of human operators. The last half-decade has seen a steep rise in the number of contributions to this area from both the control and reinforcement learning communities. Our goal is to provide a concise but holistic review of the recent advances made in safe learning control. We aim to demystify and unify the language and frameworks used in control theory and reinforcement learning research. To facilitate fair comparisons between these fields, we emphasize the need for realistic physics-based benchmarks. Our review includes learning-based control approaches that safely improve performance by learning the uncertain dynamics, reinforcement learning approaches that encourage safety or robustness, and methods that can formally certify the safety of a learned control policy.

Paper | Preprint | Talk | Slides

Simulation Benchmark for Safe Robot Learning

Abstract: In recent years, reinforcement learning and learning-based control—as well as the study of their safety, crucial for deployment in real-world robots—have gained significant traction. However, to adequately gauge the progress and applicability of new results, we need the tools to equitably compare the approaches proposed by the controls and reinforcement learning communities. Here, we propose a new open-source benchmark suite, called safe-control-gym. Our starting point is OpenAI’s Gym API, which is one of the de facto standard in reinforcement learning research. Yet, we highlight the reasons for its limited appeal to control theory researchers—and safe control, in particular. E.g., the lack of analytical models and constraint specifications. Thus, we propose to extend this API with (i) the ability to specify (and query) symbolic models and constraints and (ii) introduce simulated disturbances in the control inputs, measurements, and inertial properties. We provide implementations for three dynamic systems—the cart-pole, 1D, and 2D quadrotor—and two control tasks—stabilization and trajectory tracking. To demonstrate our proposal—and in an attempt to bring research communities closer together—we show how to use safe-control-gym to quantitatively compare the control performance, data efficiency, and safety of multiple approaches from the areas of traditional control, learning-based control, and reinforcement learning.

Preprint | Code

Upcoming Events

Workshop on Semantics for Robotics

Interdisciplinary workshop on semantics for robotics: from environment understanding and reasoning to safe interaction.

Prof. Michael Milford
Prof. Luca Carlone
Prof. Angela Dai
Dr. Oier Mees
Dr. Masha Itkina
Prof. Marco Pavone
Prof. Andrea Bajcsy
Prof. Koushil Sreenath
Dr. Federico Tombari
Dr. Manuel Keppler

July 15, 2024

CDC Learning-Based Control Invited Sessions

Advances in learning for dynamical systems.

Invited Sessions

Angela Schoellig
Sebastian Trimpe
Melanie Zeilinger
Matthias Müller

Related Resources

Talks and Discussions

Angela Schoellig, “safe-control-gym: A Unified Benchmark Suite for Safe Learning-Based Control and Reinforcement Learning,” ICRA 2022 Tutorial on Tools for Robotic Reinforcement Learning

Angela Schoellig, “Machine Learning for Robotics: Achieving Safety, Performance and Reliability by Combining Models and Data in a Closed-Loop System Architecture,” Intersections between Control, Learning and Optimization 2020

Angela Schoellig, “Safe Learning-based Control Using Gaussian Processes,” IFAC World Congress 2020

Main Contributors

Lukas Brunke
PhD Student, UofT and TUM

Adam Hall
PhD Student, UofT

Federico Pizarro Bejarano
PhD Student, UofT

SiQi Zhou
Senior Scientist, TUM

Angela Schoellig
Professor, UofT and TUM

Melissa Greeff
Assistant Professor, Queen’s University

Justin Yuan
Software Engineer, Qualcomm

Jacopo Panerati
Research Lead in ML for Control, TII


Somil Bansal, University of Southern California
Animesh Garg, University of Toronto
Matthias Müller, Leibniz University Hannover
Davide Scaramuzza, University of Zürich
Sebastian Trimpe, RWTH Aachen University
Melanie Zeilinger, ETH Zürich

University of Toronto Institute for Aerospace Studies