Reproduce “Certifiably Optimal Monocular Hand-Eye Calibration” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Emmett Wise, Matthew Giamou, Soroush Khoubyarian, Abhinav Grover, Jonathan Kelly
Researchers from the University of Toronto's STARS Laboratory developed a method for certifiably globally optimal hand-eye calibration, uniquely extending it to monocular cameras which do not provide metric scale. This work provides the first analytical proof of global optimality for hand-eye calibration algorithms in the presence of bounded sensor measurement error, showing superior accuracy over suboptimal methods in synthetic experiments.
Key insights
- Hand-eye calibration, even with unknown monocular scale, can be reformulated as a convex semidefinite program, providing a mathematical certificate of global optimality.
- The convex relaxation exhibits strong duality and stability, meaning global optimality guarantees hold even when sensor measurements contain bounded noise.
- Strict convexity of the cost function, which ensures a unique global minimum, depends on sufficient sensor motion including rotations about at least two distinct axes and corresponding translations.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.