TRIO Lab Trustworthy Robotics, Intelligence, and Optimization Lab

Global optimization for realistic robotics

...tackling non-convex, non-smooth and long-horizon problems

We push the boundaries of what optimization problems are solvable in robotics, by developing tools based on global optimization.

There is always a trade-off between the expressiveness of an optimization problem and the difficulty of solving it. To create more capable robots, we need to push the boundaries of what optimization problems we can solve. For example, explicitly modeling contact interactions instead of using predefined gates and contact sequences can lead to the discovery of more efficient and robust behaviors, using more accurate sensor and noise models leads to better estimation performance, using more accurate dynamics leads to better control performance, and so on. We are actively exploring how to efficiently solve these hard problems in estimation and control, using tools from global optimization.

A main challenge is improving efficiency of solvers and making the modeling aspect more seamless. See our projects on seamless modeling and optimization for an overview of our efforts in this direction.

Related Publications

Sampling-Based Global Optimal Control and Estimation via Semidefinite Programming
Sampling-Based Global Optimal Control and Estimation via Semidefinite Programming
Antoine Groudiev, Fabian Schramm, éloïse Berthier, Justin Carpentier, Frederike Dümbgen
Conference American Control Conference  ·  2026
On Semidefinite Relaxations for Matrix-Weighted State-Estimation Problems in Robotics
On Semidefinite Relaxations for Matrix-Weighted State-Estimation Problems in Robotics
Connor Holmes, Frederike Dümbgen, Timothy D. Barfoot
Journal IEEE Transactions on Robotics  ·  2024
Optimal Initialization Strategies for Range-Only Trajectory Estimation
Optimal Initialization Strategies for Range-Only Trajectory Estimation
Abhishek Goudar, Frederike Dümbgen, Timothy D. Barfoot, Angela P. Schoellig
Journal IEEE Robotics and Automation Letters  ·  2024
Certifiably Optimal Rotation and Pose Estimation Based on the Cayley Map
Certifiably Optimal Rotation and Pose Estimation Based on the Cayley Map
Timothy D Barfoot, Connor Holmes, Frederike Dümbgen
Journal International Journal of Robotics Research  ·  2024
Safe and Smooth: Certified Continuous-Time Range-Only Localization
Safe and Smooth: Certified Continuous-Time Range-Only Localization
Frederike Dümbgen, Connor Holmes, Timothy D. Barfoot
Journal IEEE Robotics and Automation Letters  ·  2023
A Fine Line: Total Least-Squares Line Fitting as QCQP Optimization
A Fine Line: Total Least-Squares Line Fitting as QCQP Optimization
Timothy D. Barfoot, Connor Holmes, Frederike Dümbgen
Preprint arXiv  ·  2022