Stochastic event triggered traffic flow

Övrigt konferensbidrag, 2026

This paper develops an event-triggered stochastic traffic flow model in which driver interactions are represented as state-dependent counting
processes. The resulting dynamics form a piecewise-deterministic Markov  process (PDMP), combining deterministic motion with stochastic interaction
events. Conditionally on the spacing trajectory, the interaction mechanism  admits a Poisson interpretation. A distribution-corrected interaction rate connects the microscopic  event-driven dynamics to macroscopic flow, yielding a consistent stochastic closure. Analytic properties of the diffusion approximation is presented around equilibrium flow conditions. Several numerical algorithms are proposed: closed form event generator and splitting method based numerical algorithms and they are featured against Gillespie Ogata as grand through. Numerical tests confirm that the proposed method is qualified to reconstruct not only emulated data but also real world trajectories (HighD).