it's called a "quotient space" and its elements are called "cosets" and consist of "parallel translates of a subspace by a vector".
that's why v has the overline: it's the SET:
{v+u: u in U}.
which can be thought of as the entire subspace U, moved in the direction/distance of v.
another way to think of it, is as regarding the entire space (in this case R4) losing dim(U) dimensions, by regarding all points in U as "equivalent (essentially 0)".
if dim(U) = 1, each coset is a parallel line, and you need a 3-vector to tell you "which line".
if dim(U) = 2, each coset is a parallel plane, and you need a 2 vector to tell you "which plane".
higher dimensions are harder to visualize, but the same sort of logic applies.
simce v+U is a set, v is just a "representative", and the same coset v+U can have different representatives.
one common way quotient spaces arise is in analyzing linear maps: often, we don't care about the kernel of a linear map (because everything in it just maps to the 0-vector), so we "mod it out". the resulting quotient space is isomorphic to the image space (this is pretty much equivalent to the rank-nullity theorem, but in a more abstract setting).
you calculate with elements in R4/U pretty much like you do with elements in R4, but with a "+U" along for the ride:
v+U + w+U = (v+w)+U
a(v+U) = av+U
the overline notation is a bit "cleaner" but hides some of what is going on.