While current solutions already exist on the market, some fusing data driven and simulation-based approaches like Aimsun Live, improvements on prediction quality and computational speed are necessary.
Such improvements are being developed within a separate project agreement while this particular project focuses on deploying an Aimsun Live operational pilot for Western Australia to serve as a base for these R&D activities and make sure they are applicable to real life deployment.
In comparison to the conventional offline models used for planning, (which are typically calibrated to a typical day and limited to peak periods) a solution like Aimsun Live is required to be calibrated against a much larger dataset to handle 365 days 24 hours high quality prediction.
As the difficulty in calibrating such a model increases, it is expected that Artificial Intelligence (AI) and Machine Learning (ML) techniques can assist human modelers by spotting disparity between predicted results and field observation and help identify parts of the model that require special attention. This solution would also suggest possible changes to the simulation and analytical prediction model parameters. It aims at assisting rather than replacing human modellers.
In order to facilitate the above AI and ML study (which will be completed as a separate R&D project), Aimsun has been engaged to install an Aimsun Live pilot system for Perth CBD as the testbed for the study.
The program of works will be delivered in two stages and will form several project agreements (number to be determined):
It is planned to involve other jurisdictions into the program to provide more robust and verified research into the product in a multi-state environment. Each jurisdiction will have a stage 1 pilot project when applicable and ideally, all jurisdictions will form part of stage 2 including the associated local university for each R&D topic.
Provide and install an Aimsun Live pilot system for Perth CBD. The system will provide real-time simulation-based prediction for the next 60 minutes of traffic at 15-minute intervals. The simulation-based prediction will be updated every 5 minutes.
The quality of the prediction will be assessed at the end of each hour post prediction and will be displayed by the Quality Manager in the quality control module in Aimsun Live.