OD trip forecasting using passive and readily available data

Leveraging a wealth of passive and readily available data, the proposed trip forecasting framework–developed as part of the Big data for strategic transport planning project provides robust insights into predicting Origin-Destination (OD) trips across various transport modes. The proposed approach offers a promising alternative to conventional Travel Demand Models (TDMs) for generating OD trip forecasts.

Travel Demand Models (TDMs) have long been one of the main sources of insights for strategic transport planning. The conventional approach involves dividing the study area into smaller zones, known as Traffic Analysis Zones (TAZs), and analysing the relationships between travel movements across these zones and various factors such as sociodemographic characteristics, land use patterns, spatial attributes, and network connectivity within and between the TAZs.

While TDMs are designed to help strategy developers and policymakers address a wide range of questions about the what, how, and why of travel behaviour analysis, one of the most common requirements is the generation of Origin–Destination (OD) trip matrices and flows for each transport mode. In a conventional TDM framework, these mode-specific OD matrices or flows are derived from movements between different TAZs, based on the established relationships between travel patterns and the sociodemographic, land-use, spatial, and connectivity characteristics within and between the TAZs, as discussed earlier.

One of the main inputs into conventional Travel Demand Models is the Household Travel Survey (HTS), which asks participants to provide a detailed diary of their travel behaviour over the course of one or two days. These surveys are known to be costly and, depending on the method and level of detail required, can cost $150 or more per response.

In contrast, many sources of passive and readily available data, including the Internet of Things (IoT), can provide a continuous stream of travel information for a large proportion of the population over rolling time periods, at a fraction of the cost of traditional surveys.

The Big Data for Strategic Transport Planning project developed a data fusion framework to infer mode-specific OD flows for the Greater Adelaide region. This approach integrates insights from a diverse range of transport data sources, including:

• Loop detectors at signalised intersections (SCATS)
• Bluetooth sensors (Addinsight)
• In-vehicle navigation systems (HERE Technologies)
• Public transport smartcard data (Metrocard)
• Australian Census commute flows
• GIS and open spatial datasets

This project was funded by the South Australian Minister for Infrastructure and Transport, acting through the Department for Infrastructure & Transport (DIT) South Australia; and the iMOVE Cooperative Research Centre (CRC).

Project objectives

The project’s main objectives were to:

1. Infer OD flows for car travel within the region using data from Bluetooth sensors, loop detectors, in-vehicle navigation systems and the Census;
2. Infer OD flows for public transport trips within the region using data from the public transport smartcard system (Metrocard) and the Census; and
3. Develop a Travel Demand Model based on OD inferences from IoT data.

OD inferences for car travel

The proposed framework demonstrates effectiveness in estimating car trip productions and attractions, and consequently, generating Origin–Destination (OD) car trip matrices that are comparable to those produced by the conventional survey-based Strategic Adelaide Model (SAM).

By integrating multiple data sources, the framework reduces reliance on any single dataset and helps mitigate systematic biases and missing variables. Validation against Google’s real-time traffic data indicates that the inferred OD flows achieve a high level of accuracy in reflecting observed travel times during peak periods.

OD inferences for public transport trips

The public transport smartcard ticketing system provides transport planning organisations with detailed insights into public transport usage patterns. However, these systems do not always offer a complete view of Origin–Destination (OD) flows within a study area. For example, in cities such as Adelaide, passengers are required to use their smartcard only when boarding, not when alighting. To address this limitation, various methods for inferring alighting locations have been proposed over the past two decades.

As part of this study, the proposed framework combines smartcard boarding data with mode-specific commute flows reported in the Australian Census. By fusing these complementary data sources, the framework enables the inference of complete public transport OD flows, capturing both boarding and alighting locations across the network.

Development of a trip forecasting tool based on the passive and readily available data

The IoT-derived car and public transport OD inferences were used to develop a streamlined Travel Demand Model (TDM) for predicting travel patterns in the Greater Adelaide metropolitan region.

This model responds to changes in land use, population, and transport infrastructure without the data-heavy segmentation typical of traditional models. Despite its simpler structure and reduced data requirements, the model produced highly comparable outputs to the Strategic Adelaide Model (SAM).

Predictions for the base year (2016) and future scenarios (2031 and 2041) were within 15–20% of SAM estimates for total trips and exhibited similar spatial distributions at the SA2 level. These findings demonstrate that IoT-informed models can match the performance of survey-based models while offering greater scalability and cost-effectiveness.

Conclusions

Leveraging passive and readily available travel data (including those derived from IoT technologies) provides a more cost-effective means of collecting transport usage data, and at significantly greater volumes than traditional survey-based methods.

The findings of this research indicate that a framework based on passive and readily available data can produce inferences for car and public transport OD flows that are highly comparable to those from conventional survey-based TDMs.

While household travel surveys offer more detail, such as trip purpose and the trip maker’s demographics, the findings suggest that, where mode-specific OD flows are concerned, a simplified structure relying on less detailed data can produce comparable predictions. Consequently, alternative OD trip forecasting tools with simpler specifications and fewer data requirements – which can be passive and readily available data sources – appear to perform just as well as survey-based conventional TDMs, at considerably lower cost.

Known limitations and shortcomings of the proposed framework

The proposed framework focused on the most prevalent enquiry from a conventional TDM, namely mode-specific OD flows. Accordingly, replicating the full functionality of a TDM to comprehensively address the what, why, and how questions – that a TDM is designed for – was not the intended purpose of this study.

Expected project impacts

This study shows that implementing frameworks based on passive and readily available data (including data from IoT technologies) for predicting mode-specific OD trip flows has the potential to serve as a supplementary source of insights for transport planning in cases where a lack of survey data prevents the development of conventional TDMs.

This study can also serve as a stepping stone towards developing TDM frameworks based on passive and readily available data, independent of costly surveys.

Department for Infrastructure and Transport

Transport networks are now generating more mobility data than ever before. The real challenge is not lack of information, but making sense of this information in a robust and unbiased way. Our data fusion framework demonstrates how IoT sources can be brought together to produce powerful, decision-ready insights that will shape the next generation of transport planning.

Akshay Vij, Associate Professor, University of South Australia

Final report

A final report for this project was produced, and is being used internally.