A nonlocal dispersive optimal transport: Formulation and algorithm

We propose a unified framework that effectively characterizes challenging phenomena such as anomalous transport in heterogeneous media and long-range memory effects and interactions. This framework transports agent densities from a prescribed initial distribution to a terminal distribution while minimizing the associated energy cost. Motivated by optimal transport theory, we introduce a nonlocal dispersive optimal transport (NDOT) model governed by a space–time fractional partial differential equation (PDE). We solve the NDOT formulation using the general-proximal primal–dual hybrid gradient (G-prox PDHG) algorithm, and then introduce a novel preconditioner derived from the discretization of the space–time fractional PDE to accelerate the convergence. Numerical experiments – especially those with target states represented by power functions typical of fractional differential equation solutions – show that our model substantially reduces kinetic energy costs compared with its integer-order counterparts, highlighting its effectiveness and applicability for complex phenomena such as anomalous transport in heterogeneous environments.