A Direct Calibration Framework for Gravity-Based Trip Distribution Using ICT Travel Time Data Types

Abstract

This study aims to enhance the gravity-based trip distribution calibration framework through the use and adaptation of Information and communication technologies (ICT)-based travel time data types, such as the origin–destination (O-D) travel time matrix (TTM) and partial route (PAR) travel time vector (TTV). The bi-level solution algorithm follows several key steps: obtaining the observed and instantaneous travel time values, generating demand using the doubly constrained gravity model parameters to be optimized (upper level), assigning generated demand to the network (lower level), and comparing the observed/estimated travel time results. The framework is initially evaluated with different demand scenarios (DSs) and congestion levels in the Sioux Falls (SF) test network. Subsequently, a local sensitivity analysis of initial inputs and spatial selections is conducted, followed by a numerical comparison with the traditional calibration method and conventional data source. While the framework is validated using the tests, the initial input details become increasingly important as the congestion levels deviate from the average. Finally, a real-world network (Sioux Falls, SD, USA) implementation is carried out, and the model is demonstrated to have a positive correlation with observed traffic counts and travel times. Consequently, the proposed framework offers potential benefits for Intelligent Transportation Systems (ITS).