Routing Policy Choice Models in Stochastic Time-Dependent Networks
Funding Source: UTC
Title: Routing Policy Choice Models in Stochastic Time-Dependent Networks: The Stockholm Case Study
Summary: The objective of the project is to specify and estimate a routing policy choice model using taxi GPS (Global Positioning System) traces in the city of Stockholm, Sweden. Traffic networks are inherently uncertain with random disruptions such as incidents and bad weather, and real-time information on realized traffic conditions could potentially help reduce the uncertainty and thus travelers could adjust route choices accordingly. A routing policy is defined as a decision rule that maps realized traffic conditions to actions on what link to take next at a given time and link. It represents traveler’s capability to adapt to network conditions and look ahead for information not yet available at the time of route choice decision-making. Major tasks of the project include: 1) map-matching taxi GPS readings to generate chosen routes between OD (origin-destination) pairs and the empirical time-dependent travel time distribution; 2) generating routing policy choice sets between a selected number of OD pairs and testing coverage and adaptiveness; 3) specifying and estimating a random utility model of routing policy choice; and 4) comparing the routing policy choice model with a benchmark path choice model. The project will contribute to the state of the art by estimating the first routing policy choice model using real-life data. Such a model will be valuable for the design and evaluation of traveler information systems developed by government agencies, private companies and their collaborative efforts. Additionally the team will apply the developed route choice model to another data set collected at Massachusetts to test the geographic transferability of the model. The data set contains over 100 drivers’ trajectories over a 6-month period using corporate vehicles.
Team: Song Gao