pipette
ESEspañol

Multiscale distributed MPC for Optimal Transport

Guillem Pascual, Yu-Jen Huang, Sonia Mart\'inez

Preprint

In the authors' words

In this paper, we present a novel multiscale formulation for the Optimal Transport (OT) of large swarms. The scheme is divided in two levels, a top-level OT formulation with a terminal penalty cost to handle time-varying densities, and a low-level, Model Predictive Controller (MPC) that allows agents to track the nominal velocity under control affine dynamics and rejects external disturbances. We show that the terminal-penalty dynamic formulation reduces to a regularized static dual, which agents solve in a distributed way to compute the nominal velocity.We also prove distributional input-to-state stability in the -Wasserstein metric of the density flow with respect to the mismatch that the low-level and the target motion introduce, so that the tracking error only perturbs the macroscopic flow by a bounded, quantifiable amount. Finally, we demonstrate via simulations the effectiveness of the proposed method handling time-varying targets and state perturbations.

Main resultThe abstract does not state a limitation.

Appeared: Thursday, October 15. arXiv. Preprint, not yet peer-reviewed.

Authors' comment: 9 Pages, 1 Figure, Submitted to American Control Conference 2027, under review