A multi-material transport problem with arbitrary marginals

Abstract

In this paper we study general transportation problems in ${\mathbb {R}}^n$, in which m different goods are moved simultaneously. The initial and final positions of the goods are prescribed by measures $\mu ^-$, $\mu ^+$ on ${\mathbb {R}}^n$ with values in ${\mathbb {R}}^m$. When the measures are finite atomic, a discrete transportation network is a measure T on ${\mathbb {R}}^n$ with values in ${\mathbb {R}}^{n\times m}$ represented by an oriented graph ${\mathcal {G}}$ in ${\mathbb {R}}^n$ whose edges carry multiplicities in ${\mathbb {R}}^m$. The constraint is encoded in the relation ${\mathrm{div}}(T)=\mu ^- -\mu ^+$. The cost of the discrete transportation T is obtained integrating on ${\mathcal {G}}$ a general function ${\mathcal {C}}:{\mathbb {R}}^m\rightarrow {\mathbb {R}}$ of the multiplicity. When the initial data $\left( \mu ^-,\mu ^+\right)$ are arbitrary (possibly diffuse) measures, the cost of a transportation network between them is computed by relaxation of the functional on graphs mentioned above. Our main result establishes the existence of cost-minimizing transportation networks for arbitrary data $\left( \mu ^-,\mu ^+\right)$. Furthermore, under additional assumptions on the cost integrand ${\mathcal {C}}$, we prove the existence of transportation networks with finite cost and the stability of the minimizers with respect to variations of the given data. Finally, we provide an explicit integral representation formula for the cost of rectifiable transportation networks, and we characterize the costs such that every transportation network with finite cost is rectifiable.