Modelling Ridesharing in a Large Network with Dynamic Congestion

In ridesharing, commuters with similar itineraries share a vehicle for their trip. Despite its clear benefits in terms of reduced congestion, ridesharing is not yet widely accepted. We propose a specific ridesharing variant, where drivers are completely in exible. This variant can form a competitive alternative against private transportation, due to the limited e orts that need to be made by drivers. However, due to this in exibility, matching of drivers and riders can be substantially more complicated, compared to the situation where drivers can deviate. In this work, we identify the e ect of such a ridesharing scheme on the congestion in a real network of the ^Ile-de-France area for the morning commute. We use a dynamic mesoscopic trac simulator, Metropo- lis, which computes departure-time choices and route choices for each commuter. The matching is solved heuristically outside the simulation framework, before departures occur. We show that even with in exible drivers, ridesharing can alleviate congestion. By slightly increasing exibility, the performance of the ridesharing scheme can be further improved. Furthermore, we show that ridesharing can lead to fuel savings, CO2 emission reductions and travel time savings on a network level, even with a low participation rate.