KernelSOS

Kernel-based Sum-of-Squares Optimization for Robotics

KernelSOS is a kernel-based sum-of-squares optimization framework for robotics applications (Groudiev et al., 2025).

Illustration of the principle of Kernel Sum-of-Squares Optimization. A polynomial nonnegativity constraint is relaxed to a kernel-based constraint, leading to a scalable optimization framework.

References

  1. arXiv
    Sampling-Based Global Optimal Control and Estimation via Semidefinite Programming
    Antoine Groudiev, Fabian Schramm, Éloïse Berthier, Justin Carpentier, and Frederike Dümbgen
    arXiv preprint, 2025
    Best Poster Award PDF Poster Slides