A slow-to-start traffic model related to a M/M/1 queue

Abstract

We consider a system of ordered cars moving in $\R$ from right to left. Each car is represented by a point in $\R$; two or more cars can occupy the same point but cannot overpass. Cars have two possible velocities: either 0 or 1. An unblocked car needs an exponential random time of mean 1 to pass from speed 0 to speed 1 (\emph{slow-to-start}). Car $i$, say, travels at speed 1 until it (possibly) hits the stopped car $i-1$ to its left. After the departure of car $i-1$, car $i$ waits an exponential random time to change its speed to 1, travels at this speed until it hits again stopped car $i-1$ and so on. Initially cars are distributed in $\R$ according to a Poisson process of parameter $\lambda<1$. We show that every car will be stopped only a finite number of times and that the final relative car positions is again a Poisson process with parameter $\lambda$. To do that, we relate the trajectories of the cars to a $M/M/1$ stationary queue as follows. Space in the traffic model is time for the queue. The initial positions of the cars coincide with the arrival process of the queue and the final relative car positions match the departure process of the queue.