Researching the Influence of Time-dependent Tolls with a Multi-agent Traffic Simulation
M Rieser, U Beuck, K Nagel, VSP, TU-Berlin TU-Berlin, DE
We research the influence of time-dependent tolls on travellers using a multi-agent transport simulation, especially the influence a toll can have outside its range of time the toll has to be paid.
Most existing macroscopic transportation planning tools are able to include static tolls in their static traffic-flow assignments by adjusting costs or attractiveness of single links or routes. But they usually fall short when modelled tolls should be time- or demand-dependent (e.g. the toll only needs to be paid during rush hour or when the amount of vehicles on a link is beyond a predefined limit).
Some macroscopic tools offer dynamic assignment of traffic and would be able to represent time-dependent tolls in their models. Routing reactions to such tolls could thus be represented. But these models still have problems with traveler reactions in the time direction -- even if trips in a given time slice can be suppressed according to some elasticity measure, they usually do not show up at other times in these models. The only exception that we are aware of is METROPOLIS (de Palma and Marchal), which self-consistently selects routes AND departure times for trips.
Our approach goes one step further and models travellers' full daily plans. As a result, reactions in the time direction are extended further. A toll only active during the evening rush hour would not only lead to people leaving work either before or after the toll has to be paid, but people would plan to begin work in the morning earlier or later, depending on their planned departure in the evening.
Technically, we value travellers' daily plans by a utility function, which is a sum of
- positive utility (activity-specific) that is earned when doing an activity
- negative utility for traveling
- negative utility for being early or late
Because of its formulation, the approach is able to automatically take care of constraints, such as the one described above, or the necessity to leave home earlier in the morning in order to get the shopping done after work and before shops close. The utility functions are obtained from value-of-time estimations by Axhausen and others.
Existing tools are not able to model such relations, as every trip is modelled on its own and not in the context of one traveller with the knowledge of earlier or later trips the same day. Multi-agent traffic simulations can have agents with their complete day-plans and are thus able to better model the influence of time-dependent tolls.
We research the influence of different tolling schemes on daily commuters to and from the city of Berlin, Germany, using our implementation of a multi-agent transport simulation (MATSim). Especially, we are interested how agents react to time-dependent tolls, and how strong the influence of a toll during the evening rush hour is on the morning rush hour. We further research how bottlenecks in the road network behave when a time-dependent toll is applied to them and how the observed traffic pattern at tolled bottlenecks can be brought in context with the economics of the bottleneck described by Arnott, de Palma and Lindsey.
Association for European Transport