We’re now entering Stage 2 of this transport predictive solution for Perth, WA, focused on artificial intelligence for calibration against much larger real-time datasets. The WA node will focus on developing and testing improved model calibration capability for both live and offline transport models, ensuring prediction accuracy for any hour of the day, seven days a week.
We’re partnering with iMOVE Australia, Main Roads Western Australia, and The University of Western Australia.
This project aims to offer a real-time decision support tool for traffic operations centres to predict congestion on the network, quickly assess the impact of unplanned events and evaluate the mitigation potential of several possible responses.
Such a solution will help reduce congestion, especially in non-recurrent situations, and significantly increase travel time reliability.
The use of tools to facilitate longer-term prediction of how transportation networks will perform in the future is a well-established practice in strategic planning by transport authorities. Tools to support day-to-day operations, relying on short-term predictions, are in their infancy, especially in Australia.
Particular objectives to enhance short-term prediction performance are:
- Smart sensing for enhanced travel demand estimation; and
- Artificial intelligence (AI) and machine learning (ML) for calibration against much larger real-time datasets
The WA node will focus on (2) developing and testing improved model calibration capability for both live and offline models, ensuring prediction accuracy for any hour of the day, seven days a week.
This research proposes to improve model calibration and the accuracy of 24 hour/ 7-day models (live and offline) for not just the AM and PM peaks but any hour of any day. The research results will be tested in a WA Aimsun Live network pilot model, developed as part of the more comprehensive project. Further evaluation and performance accessibility of tools developed in this research will be performed in QLD Aimsun Live network model.