Dynamic Origin-Destination Estimation Using Smart Card Data: An Entropy Maximisation Approach

TL;DR

This paper presents a novel entropy maximisation approach for dynamic origin-destination estimation using smart card entry data, effectively solving large network problems with Lagrangian relaxation and achieving accuracy comparable to manual surveys.

Contribution

It introduces an entropy maximisation model for OD estimation with only station-entry data and demonstrates its effectiveness on a real Stockholm case study.

Findings

Model achieves comparable accuracy to manual surveys.

Using additional data improves estimation accuracy.

Lagrangian relaxation efficiently solves large-scale problems.

Abstract

Problems of dynamic origin-destination (OD) estimation using smart card data can be modelled using entropy maximisation and solved for large networks using solution techniques such as Lagrangian relaxation. In this paper, we give an overview of the research literature about OD estimation. We show how entropy maximisation can be used to model this problem in case station-entry data from smart cards is the only available information, i.e. number of entries is known but not exits nor the flow between the stations. The large entropy maximisation program is solved using Lagrangian relaxation. The model is tested on a case study from the commuter train service in Stockholm with smart card entry-data from a working day in 2015. The results show that, given the entry-data, the entropy maximisation-based model and methods allows to find an OD-estimate that is as accurate as the reported estimates from manual observations. The estimates are further improved when using relevant assumptions and additional available data.