Our article includes tutorials on three classes of safety filters: Hamilton-Jacobi Reachability, Control Barrier Functions, and Predictive Control. The article covers applications to first-principle and data-driven models, a selection of model-based and learning-based controllers, and several illustrative case studies from the robotics domain.
By collaborating with experts in the areas of Control Barrier Functions (Andrew Taylor, Prof. Aaron Ames, Caltech) and Safety Filters using Hamilton-Jacobi Reachability Analysis (Jason J. Choi, Prof. Koushil Sreenath, Prof. Claire Tomlin, UC Berkeley), the article aims to provide a comprehensive overview of the state of the art, including work on predictive safety filters done in my former group ICS – Intelligent Control Systems at ETH Zurich (Prof. Melanie Zeilinger).
Our recent post on the Bosch Research Blog showcases some exciting future applications for safety-critical control methodologies and their interplay with humans and ai.
Two new pre-prints are available, which will be presented at the European Control Conference 2023:
- Approximate Predictive Control Barrier Functions using Neural Networks: A Computationally Cheap and Permissive Safety Filter
- LQG for Constrained Linear Systems: Indirect Feedback Stochastic MPC with Kalman Filtering
We are excited that our paper “Cautious Bayesian MPC: Regret Analysis and Bounds on the Number of Unsafe Learning Episodes” was accepted and will be published in the IEEE Transactions on Automatic Control in August 2023.
We have contributed to the session “A Tutorial on Nonlinear Model Predictive Control: What Advances Are On the Horizon?” at the American Control Conference, Atlanta, 2022. The corresponding paper will be published in the conference proceedings and can be found, e.g. here.
I am happy to share that our work on connecting predictive safety filters and control barrier functions, “Predictive control barrier functions: Enhanced safety mechanisms for learning-based control” will be published in the IEEE Transactions on Automatic Control in May 2023.
Our paper Learning-based Moving Horizon Estimation through Differentiable Convex Optimization Layers has been accepted for the 4th Annual Learning for Dynamics & Control Conference at Stanford and was selected for oral presentation.
Showcasing some of our safety-critical control methods at Bosch Research
Spotlight talk at IROS 2021 Workshop on Safe Real-World Robot Autonomy: Summarizes some of the more recent work on predictive safety filters.
A predictive safety filter for a miniature racing application and its combination with imitation learning to safely learn an expert policy: Preprint
I’m working at Bosch Research on the safe control of uncertain dynamical systems. Before, I was a postdoctoral researcher in the intelligent control systems group at IDSC ETH. My research interests are Safe Learning Control, Model Predictive Control, and Model-based Reinforcement Learning; also, see my Google Scholar profile.
For a detailed CV, see my LinkedIn profile.