tashirosgt wrote:
Find an efficient set of constraints that determine whether an axis-aligned rectangular solid is inside a polyhedron. Be more economical than writing constraints that say each corner of the rectangular solid is inside it.
I found a solution to this problem on the web, but there was no proof of the answer. Can anyone explain why the following constraints would work?
Let

and

be

dimensional vectors with

( The "

" means a component-by-component relation.) Let

be the rectangular solid defined by the vectors

(So

is the "lower corner" and

is the "upper" corner). Let

be the polyhedron defined by the set of vectors

where

is a given matrix with

row entries and

is a given vector of the same column dimension as

.
Define

to be the matrix with entries

.
Define

to be the matrix with entries

.

iff

I think you want

instead of the other way round (otherwise your cuboid is empty).
The left hand side of each constraint

is the largest possible

as v ranges over the vertices

, by the choice of A^+ and A^- (if the coefficient is negative, choose the lower so it subtracts less, and if the coefficient is positive, choose the upper so it adds more).