An improved replanning strategy for congested traffic conditions in MATSim

Agent-based transport models such as MATSim have been developed and used for the past few decades to simulate transportation systems in different cities and are capable of accurately reproducing travel patterns at different levels of aggregation. Given their disaggregate representation of both travel demand and supply as well as the high level of spatial and temporal detail, agent-based transport models also have the potential to allow for unprecedentedly detailed and disaggregated analysis. When simulating large study areas, it is often common to increase the replanning rate to achieve a quicker convergence. However, this comes at the cost of causing substantial oscillations in link-level dynamics which render disaggregate analysis difficult, if not impossible. This paper tries to reconcile both objectives: favour quicker convergence as with higher replanning rates while mitigating the strong oscillations of the link-level dynamics.