arXiv subjects

Xintian Liu

Publications and source records attributed to Xintian Liu.

3 recordsLinked to original sources

arXiv2026

Master Equations for Mean Field Game of Controls: A Unification of Weak Solution Notions

In this manuscript, we study the master equation for mean field game of controls, assuming only that the data are Lipschitz continuous in the measure variable. Accordingly, we propose a weaker notion of solution called weak solutions for the master equation. We establish the global well-posedness of the master equation for such solution under the Lasry-Lions monotonicity and displacement $λ$-monotonicity conditions, respectively. The arguments rely on the analysis of the associated Pontryagin forward-backward stochastic differential equation system, especially the stability in the measure variable. Finally, we review several existing notions of non-smooth solution from the literature, namely good solutions, weak viscosity solutions, Lipschitz solutions, monotone solutions, and examine their relationship with our weak solution. We show that under appropriate assumptions all of them are equivalent by their definitions, and thus our manuscript provides a unification of weak solution notions for the master equations derived from mean field game of controls.

math.OC↗

arXiv2026

OhmicFlow: Forecasting transit passenger flow under extreme weather disruptions via Ohm's law

As climate change intensifies, extreme weather events (EWEs) increasingly disrupt the supply-demand balance of mass transit systems. Accurate and reliable prediction of origin-destination (OD) passenger flow is essential for timely emergency response, but remains difficult under abnormal conditions. Although existing physics-informed methods can enhance model robustness, they remain insufficient for transit systems under extreme weather disruptions, where passenger flow fluctuations are jointly shaped by the redistribution of latent demand as a direct result of EWEs (i.e., direct effects) and the increase in travel impedance due to EWE-induced supply contraction and congestion effects (i.e., indirect effects). To address these limitations, we first conceptualize the transit network as an electrical circuit, treating latent demand as voltage, travel impedance as resistance and passenger flow as current, and then propose the OhmicFlow framework to jointly predict these variables under disruptions through Ohm's law. Specifically, a future-aware spatiotemporal backbone is used as an ammeter to predict disrupted passenger flow, and its replica is reused as a bypass voltmeter with impedance controlled in the inputs to infer latent demand counterfactually and in parallel. Travel impedance under foreseeable EWE disruptions is further modeled using a thermistor analogy, enabling dynamic estimation by coupling supply contraction with congestion effects. A multi-objective loss is incorporated to fit observed data while enforcing the Ohmic constraint. Empirical experiments based on 10 years of Shenzhen Metro data covering 17 EWEs show that OhmicFlow consistently outperforms various baseline methods, achieving lower prediction errors across various chronological training settings while improving uncertainty calibration, robustness, transferability, and interpretability.

physics.soc-ph↗

arXiv2026

The global well-posedness for master equations of mean field games of controls

In this manuscript, we establish the global well-posedness for master equations of mean field games of controls, where the interaction is through the joint law of the state and control. Our results are proved under two different conditions: the Lasry-Lions monotonicity and the displacement $λ$-monotonicity, both considered in their integral forms. We provide a detailed analysis of both the differential and integral versions of these monotonicity conditions for the corresponding nonseparable Hamiltonian and examine their relation. The proof of global well-posedness relies on the propagation of these monotonicity conditions in their integral forms and a priori uniform Lipschitz continuity of the solution with respect to the measure variable.

math.PR↗