Maximum-Volume Cuboid
Find the maximum-volume axis-parallel cuboid inscribed in a convex polyhedron .
This example demonstrates how optimization of a product of positive terms can be expressed in terms of power-cone constraints that can be used with ConicOptimization to find the optimum.
Consider a random convex polyhedron constructed as the convex hull of random points.
For the cuboid, find a lower-corner point and a vector of side length
so that the cuboid is represented in the Wolfram Language by
. The volume of the cuboid is just the product of the side lengths, so the objective is to maximize
. If all of the corners of the cuboid are in
, then all of the points in the cuboid are also. The corners may be described by
, where
is in the set
of all possible n‐tuples of elements from
.
The problem becomes:
Since is non-negative, maximizing the product
is the same as maximizing the geometric mean
, which is known to be concave. Maximizing
is equivalent to minimizing
, which is convex. Using an auxiliary variable
, reformulate the problem with a linear objective function -
with the additional constraints
.
The problem becomes:
The power cone is the set of such that
, and may be expressed in the Wolfram Language by
.
Since , the new constraint
can be satisfied for non-negative
and is equivalent to
. This can be written as a series of power cone constraints
For the problem becomes:
A convex polyhedron can be represented as intersections of half-spaces . Extract the coefficients
for each side.
Solve the problem.
Show the maximum-volume-inscribed cuboid.
Instead of a polyhedron, take any convex conic representable set K⊆n—for example, an ellipsoid. A vertex of the cuboid is inside the ellipsoid iff
.
Solve the problem.
Plot the result.