An ellipsoid could be rolled (without slippage) on a horizontal plane so that its point
of contact traces out a closed geodesic on its surface:
Q1. Which other convex bodies (in $\mathbb{R}^3$) share this property?
A version of this question was raised in an earlier MO question
("Rolling a convex body: Geodesics vs. rolling curves")
in which Andrey Rekalo pointed to the book
Geometry of nonholonomically constrained systems
by Richard H. Cushman, Hans Duistermaat, and Jędrzej Śniatycki.
This book indeed describes the relevant equations of motion, but I am
not
finding it easy to extrapolate from their discussion to answer the
posed question.
Q2. What conditions suffice to guarantee that
a particular closed geodesic is the trace of a rolling curve?
In the case of the ellipsoid, the normal to the point of rolling
contact aims at all times through the center of gravity.
Perhaps it suffices that the center of gravity lies in the plane
determined by that normal and the tangent to the rolling curve.
I was particularly wondering:
Q3. Will a Zoll surface, all of whose geodesics
are closed, roll along (some of its) geodesics?
It would be quite interesting if the answer is 'Yes', although I think
this is unlikely.
(Zoll surfaces were discussed in two earlier MO questions:
"Surfaces all of whose geodesics are both closed and simple"
and
"Riemannian surfaces with an explicit distance function?.")
Thanks for any suggestions or pointers!
Best Answer
I don't have a definitive answer, but surely any convex surface $S$ in 3-space that has reflectional symmetry about a plane $P$ will roll along the geodesic $S\cap P$. If you take such a surface with 120-fold icosohedral symmetry group in $O(3)$, you'll get quite a few such geodesics, more than the (generic) ellipsoid has.