Slow-to-Start Traffic Model: Condensation, Saturation and Scaling Limits

Abstract

We consider a one-dimensional traffic model with a slow-to-start rule. The initial position of the cars in $\mathbb R$ is a Poisson process of parameter $\lambda$. Cars have speed 0 or 1 and travel in the same direction. At time zero the speed of all cars is 0; each car waits an exponential time to switch speed from $0$ to $1$ and stops when it collides with a stopped car. When the car is no longer blocked, it waits a new exponential time to assume speed one, and so on. We study the emergence of condensation for the saturated regime $\lambda>1$ and the critical regime $\lambda=1$, showing that in both regimes all cars collide infinitely often and each car has asymptotic mean velocity $1/\lambda$. In the saturated regime the moving cars form a point process whose intensity tends to 1. The remaining cars condensate in a set of points whose intensity tends to zero as $1/\sqrt t$. We study the scaling limit of the traffic jam evolution in terms of a collection of coalescing Brownian motions.