An Agent-based Simulation Model of Traffic Congestion
J Mc Breen, P Jensen, LET & ENS-Lyon, F Marchal, LET, FR
We present an agent-based exploratory study of a simple congestion
model. The forms and levels of heterogeneity required for the model
to be stable are investigated.
We present an agent-based exploratory study of a simple congestion model. This model is based on the pioneering work of Vickrey, who examined the effect of departure time choice on congestion for a single road bottleneck. The agent dynamics, followed in order to reduce the total cost of the journey, including congestion costs and the cost of not arriving at the desired time, are scalable to city sized models.
Each simulated agent represents an individual and was given a simple behavioural rule which is followed in order to reduce the cost of the
journeys undertaken. An agent who reviews his departure time calculates the cost of a randomly chosen departure time, and changes
if this cost is sufficiently cheaper than his current cost.
Homogeneous agent systems, in which all agents wish to arrive at the same time, are unstable. This casts doubt on the stability of the equilibrium of the theoretical bottleneck model. The proportion of agents who review their departure times in the same iteration effects the stability of the system. When the agents were given a normal distribution of preferred arrival times the system was stabilised while the level of congestion remained significant. The variance of this distribution and the reviewing rate are two important parameters that determine the qualitative behaviour of the model. A graph of the
stability of the system against these two parameters highlights the
parameter space of stable behaviour.
The introduction of agents with different propensities to review their departure times is found to have a slight stabilising effect. Simulations performed with agents who had a distribution of sensitivities to schedule delays showed significant instability.
Association for European Transport