A hands-on tutorial on learning-to-optimize methods for real-time MPC, sampling-based control, and adaptive robotic control.
This tutorial introduces learning-to-optimize (L2O) as a practical framework for accelerating and improving optimization-based control in robotic applications.
The tutorial focuses on real-time model predictive control, efficient sampling-based control, and learning-to-adapt methods for controllers operating under changing dynamics and uncertain environments. The hands-on examples will use the Ctrl-L2O toolbox to demonstrate these ideas across robotic systems such as autonomous vehicles, quadrupedal robots, and bipedal robots, with interfaces compatible with platforms such as MuJoCo.
| Time | Speaker | Talks |
|---|---|---|
| 0:00-1:00 | Viet-Anh Le | Learning-to-Optimize and Its Applications in Control |
| 1:00-1:30 | — | Coffee Break |
| 1:30-2:30 | Yorie Nakahira | (TBA) |
| 2:30-3:00 | — | Coffee Break |
| 3:00-4:00 | Speaker 3 (TBA) | (TBA) |
This talk introduces learning-to-optimize (L2O) through the lens of parametric optimization for robotic control. It covers model predictive control (MPC), model predictive path integral control (MPPI), and core L2O formulations, including learned solution maps, warm starts, algorithm unrolling, and differentiable optimization.
Speaker biography to be announced.
(TBA)
Speaker biography to be announced.
Speaker biography to be announced.
Slides, software toolboxes, and notebooks for the talks will be posted before the tutorial.
Organizer names, affiliations, and contact information will be added here.
This tutorial is being developed for the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) 2026. Additional acknowledgments will be added as the tutorial materials are finalized.