[Paper Review] The deterministic-stochastic flow model

Alexander P. Buslaev, В. М. Приходько+2 more|ArXiv.org|Apr 20, 2005

Simulation Techniques and Applications3 references17 citations

TL;DR

This paper introduces the Deterministic-Stochastic Traffic (DST) flow model, which integrates deterministic vehicle dynamics with stochastic lane-changing behavior to predict traffic flow intensity. By modeling individual vehicle maneuvers probabilistically and coupling them with dynamic spacing functions, the model derives a generalized flow-density-velocity relationship that captures instability and capacity drops under varying road conditions and traffic densities.

ABSTRACT

A discrete model of traffic on a multilane road is considered. The traffic is presented as particles movement with a deterministic component and a stochastic one. Formulas for the traffic characteristics have been found. The model can explain theoretically some phenomena that have been discovered earlier on the basis of the experimental results.

Motivation & Objective

• To develop a unified model that captures both collective (deterministic) and individual (stochastic) behaviors in traffic flow.
• To analyze how stochastic lane changes affect flow intensity and stability, especially under high density and varying road conditions.
• To quantify the impact of fast-moving vehicles (e.g., emergency vehicles) on overall traffic flow using a perturbation framework.
• To generalize the fundamental diagram of traffic flow by incorporating probabilistic driver behavior and dynamic spacing.

Proposed method

• Models vehicle movement on multilane roads using a discrete cell-based grid synchronized across lanes.
• Defines dynamic distance $ d(v) = c_0 + c_1 v + c_2 v^2 $, where $ c_0 $ is vehicle length, $ c_1 $ reflects reaction time, and $ c_2 $ depends on road surface (e.g., dry, wet, icy).
• Introduces a stochastic transition probability $ p(r,T,v) $ for lane changes, modeled as $ p(r,T,v) \simeq p(0,T,v) R_i(r) $, with $ R_i(r) $ derived from geometric probability for $ i $-lane configurations.
• Derives average flow intensity $ \bar{q}_i = \rho v + \rho d(v) p(v) R_i(\rho d(v)) $, where $ \rho $ is lane density and $ p(v) $ is the limit of $ p(0,T,v)/T $ as $ T \to 0 $.
• Uses the inverse of dynamic distance to express flow rate in terms of velocity and density, enabling analysis of flow stability and capacity drops.
• Analyzes the impact of fast vehicles ("blue lights") by modeling mixed flows with two vehicle types and deriving a relative flow rate $ Q(\rho,v) $.

Experimental results

Research questions

• RQ1 How does stochastic lane changing behavior affect the fundamental flow-density-velocity relationship in multilane traffic?
• RQ2 What is the critical density at which flow instability or capacity drop emerges due to increased stochastic transitions?
• RQ3 How does the presence of fast-moving vehicles (e.g., emergency vehicles) alter the overall flow rate in a mixed traffic stream?
• RQ4 How do different road surface conditions (dry, wet, icy) influence the dynamic distance function and thus the flow characteristics?
• RQ5 What is the analytical form of the generalized fundamental diagram that includes both deterministic and stochastic components?

Key findings

• The model generalizes the classical fundamental diagram by incorporating stochastic lane changes, yielding $ \bar{q}_i = \rho v + \rho d(v) p(v) R_i(\rho d(v)) $, which captures flow instability and capacity drops.
• For a three-lane road at $ \rho = 0.05 $ vehicles/m, the model predicts a pronounced instability near 25 m/s (90 km/h), indicating potential for spontaneous congestion.
• On dry asphalt with $ d(v) = 5.7 + 0.504v + 0.0285v^2 $, the model shows that flow rate peaks at moderate speeds and declines sharply at high densities due to increased stochastic resistance.
• The presence of fast vehicles reduces the effective headway, and the relative flow rate $ Q(\rho,v) $ drops significantly when $ \rho d(v)/2 > 1 $, indicating instability in mixed flows.
• The model predicts that for $ v < -\frac{c_1}{2c_2} + \sqrt{\left(\frac{c_1}{2c_2}\right)^2 - \frac{c_0}{c_2} + \frac{2}{c_2 \rho}} $, the system remains stable under perturbations from fast vehicles.
• For a single lane, the flow rate $ \bar{q}_1(v,p) $ reaches a maximum at intermediate velocities, and the model shows that $ \bar{q}_1 \approx \frac{rv}{d(v)} + \frac{rp(0,T,v)}{T} $, with $ p(0,T,v)/T \to p(v) $ as $ T \to 0 $.