Investigation of Macroscopic Fundamental Diagrams in Urban Road Network using an Agent-based Model

Abstract

Congestion pricing is considered an effective management policy to reduce traffic congestion in transportation networks. To study congestion pricing schemes in urban networks, one would require a simulation model which can represent demand elasticity and traffic supply dynamics. Two main inadequacies exist in current simulation models. Firstly, the traditional traffic simulators with car-following lane-changing and route choice models consider traffic demand as input, i.e. inelastic to level of congestion conditions. Secondly, in traditional congestion pricing models with elastic demand, the utilized network supply curve is not consistent with the physics of traffic and dynamics of congestion and queues. Also, many of these models are assuming deterministic and homogeneous population characteristics. This might result in nonoptimal estimated tolls. Agent-based models are possible solutions for representing demand elasticity. This is because heterogeneous travelling agents are used in the models, (i) each agent has an individual utility function, (ii) each agent has individual value of travel time savings and (iii) one agent’s behaviour affects other agents’ decisions. For network supply modeling, it has been broadly shown through field tests and traffic simulations that traffic in large urban regions can be modelled dynamically at an aggregate level, as expressed by a Macroscopic Fundamental Diagram (MFD). If the output of an agent-based model shows the property of the MFD, it would be interesting to develop a dynamic network-wide congestion pricing schemes controlled by this macroscopic tool. Therefore the goal of this paper is to investigate whether an agent-based simulation model produces results consistent with the physics of traffic and whether a MFD can be observed. Swiss Transport Research Conference May 11 ‐13, 2011 2 Several case studies are done on Zurich urban road network in the multi agent-based traffic simulator MATSim. Results show that the productions of MATSim are consistent with the physics of traffic flow at both microscopic and macroscopic levels. Besides, MATSim is able to reproduce similar traffic phenomena such as network hysteresis loops which have been observed in previous work from real experiments and traffic simulations.