arXiv Science⌕ Search
Home/Archive/arXiv

arXiv · 2610.02537

Second-Order Mean-Field Schrödinger Bridges with Partial State Information

Asmaa Eldesoukey·Md Zulfiqur Haider·Italo Napolitano·Yongxin Chen·Abhishek Halder
View original source ↗Read source document ↗Export metadata ↓

Abstract

Motivated by scenarios in control systems where decision making is informed by only partial information about the states, we introduce a novel formulation of the Schrödinger bridge problem for inertial interacting-agent systems. Within our setting, the agent ensemble is steered to match partial specifications of the collective configuration. Our analysis pertains to a mean-field framework in which the collective dynamics are described by a nonlinear Vlasov--Fokker--Planck integro-PDE with nonlocal interactions. We assume that only distributions over positions are specified at the endpoints of a finite time horizon, representing aggregate population snapshots, while the corresponding velocity degrees of freedom remain unspecified. We derive the associated optimality system, identifying the optimal controlled evolution as jointly informed by the initial and terminal specifications through forward and backward propagation in time. We also show that the partial endpoint specifications bring about distinct conditions on the optimal control and the consistent evolution. Subsequently, to compute solutions to the optimality system, we develop a nested fixed-point algorithm inspired by the classical Fortet--Sinkhorn iteration, but extended to handle the nonlinearities arising from nonlocal interactions and temporal coupling.

Explore related subjects

math.OC

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Asmaa Eldesoukey, Md Zulfiqur Haider, Italo Napolitano, Yongxin Chen, Abhishek Halder. 2026-10-01. Second-Order Mean-Field Schrödinger Bridges with Partial State Information. https://arxiv.org/abs/2610.02537

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

arXiv2026

Asymptotic Optimality of Inventory Projection in Dual-Sourcing Systems

We consider a single-echelon inventory system under periodic review with two suppliers facing stochastic demand, where excess demand is backlogged. The expedited supplier has a shorter lead time than the regular supplier but charges a higher unit price. We introduce the Projected Expedited Inventory Position (PEIP) policy. We show that this policy is asymptotically optimal in two regimes of practical interest: (i) when the lead time of the regular supplier becomes large and (ii) when both the shortage cost and the cost premium for expedited units become large, with their ratio held constant. We show through an extensive numerical investigation that the PEIP policy outperforms the current best performing heuristic policies in literature.

math.OC↗
arXiv2026

Variance-reduced accelerated methods for decentralized stochastic double-regularized nonconvex strongly-concave minimax problems

In this paper, we consider the decentralized, stochastic nonconvex strongly-concave (NCSC) minimax problem with nonsmooth regularization terms on both primal and dual variables, wherein a network of $m$ computing agents collaborate via peer-to-peer communications. We consider when the coupling function is in expectation or finite-sum form and the double regularizers are convex functions, applied separately to the primal and dual variables. Our algorithmic framework introduces a Lagrangian multiplier to eliminate the consensus constraint on the dual variable. Coupling this with variance-reduction (VR) techniques, our proposed method, entitled VRLM, by a single neighbor communication per iteration, is able to achieve an $\mathcal{O}(κ^3\varepsilon^{-3})$ sample complexity under the general stochastic setting, with either a big-batch or small-batch VR option, where $κ$ is the condition number of the problem and $\varepsilon$ is the desired solution accuracy. With a big-batch VR, we can additionally achieve $\mathcal{O}(κ^2\varepsilon^{-2})$ communication complexity. Under the special finite-sum setting, our method with a big-batch VR can achieve an $\mathcal{O}(n + \sqrt{n} κ^2\varepsilon^{-2})$ sample complexity and $\mathcal{O}(κ^2\varepsilon^{-2})$ communication complexity, where $n$ is the number of components in the finite sum. All complexity results match the best-known results achieved by a few existing methods for solving special cases of the problem we consider. To the best of our knowledge, this is the first work which provides convergence guarantees for NCSC minimax problems with general convex nonsmooth regularizers applied to both the primal and dual variables in the decentralized stochastic setting. Numerical experiments are conducted on two machine learning problems. Our code is downloadable from https://github.com/RPI-OPT/VRLM.

math.OC↗
arXiv2026

A Data-Driven Linear Programming Model for Energy-Optimal Metro Timetables

We propose a data-driven linear programming model to compute energy-optimal timetables for networks operating under communications-based train control (CBTC); the model was developed with Hitachi Rail Canada, the largest provider of CBTC systems worldwide. Our model minimizes the network's effective energy consumption, defined as the total traction energy minus the regenerative braking energy transferred from braking trains to nearby accelerating trains. Our main modeling contribution is to show the signed overlap between a braking and an accelerating phase is concave in the event times, so maximizing the transferred energy is a convex problem that can be reformulated into a linear program via hypograph constraints. In contrast, prior work either captures this synchronization structure with binary variables or other nonconvex constraints, or relies on multiple stages interleaved with simulation. Our model computes an energy-optimal timetable subject to the operational constraints of the railway network in under a second on a laptop for every instance studied and predicts the effective energy consumption of the network without requiring time-consuming simulations. We demonstrate the effectiveness of our model on two real-world CBTC networks. First, on 11 full-day operational instances of Shanghai Metro Line 8 with 1,000-1,332 active trains, the model computes energy-optimal timetables in under 0.7 seconds per instance and predicts 19.5%-28.2% reductions in effective energy consumption relative to the existing operational timetables; the industrial physics-based simulator SPSIM independently corroborates these reductions. Second, on 17 real-world instances of the Docklands Light Railway with 108-336 active trains, the model predicts reductions of 19.3%-28.8%, solving each instance in under 0.11 seconds. Our model is being integrated into Hitachi Rail Canada's industrial timetable compiler.

math.OC↗