This project will develop a pilot model that utilises secondary datasets (e.g. signal timing data) within Main Roads Western Australia to estimate overall delay at intersections in real-time.
Real-time information, especially delay time at intersections, is valuable for traffic operations but is not readily available and costly to procure. Existing data sources that Main Roads has access to do not currently provide this information at a useful level of accuracy.
Such a model would allow Main Roads to determine the delay at a network, intersection, or at an approach level, while not requiring any additional sensor equipment or expensive data licensing agreements. It would inform decisions relating to network operational strategies and road project development.
Real-time information, especially delay time at intersections, is valuable for traffic operations but is not currently readily available. This project aims at creating a real-time model for estimating delays at Perth’s traffic signals, which would inform project decisions and operational strategies.
It has the following areas of significance:
Improving real-time information
Real-time information is limited and costly. The existing data sources that Main Roads has access to do not currently provide this information at a reasonable level of accuracy, and it can be costly to install equipment to collect the data or procure such data from third parties.
- Historical GPS data is currently captured for network reporting, but real-time information would require a large annual cost to subscribe to particular data sources. The number of samples is also limited, which impacts the overall accuracy of data.
- Some other data sources can provide near real-time information, but accuracy is a challenge.
- By using other available sources of data within Main Roads, such as signal timing data, a model for the network could be created to understand the overall delay at each intersection in real-time. This could allow Main Roads to measure the delay at a network, intersection, or approach level, while not requiring any additional expense in data licensing agreements.
Reducing human resource cost
- Modelling is costly and time consuming. With over 1,000 signalised intersections on the network, it is difficult to track, maintain and optimise this infrastructure.
- An accurate, real-time model of the network could inform decisions by our network operators, who could be better informed to make optimisation decisions, rather than spending additional time on the network. This would also improve real-time operations, by warning operators of particular incidents and issues on the network early and allow for more informed decisions.
Improving Perth’s journeys and reducing impacts
- By further optimising the efficiency of the road network, the need for investment in additional road infrastructure could be reduced.
- A reduction in overall delay would also reduce vehicle-operating costs for general traffic and public transport, in turn also reducing the network’s impact on the environment by reducing overall vehicle emissions caused by congestion
Enabling Artificial Intelligence
- Artificial intelligence algorithms use inputs and outputs in order to learn the data and produce outputs when given new inputs. From using our existing data, we could use our existing data to align with our Key Performance Indicators, which could dramatically change how we manage and operate the network.
- This could also extend a recent iMOVE – PATREC project, Improved network performance prediction through data-driven analytics and simulation. This project identified that, by developing a machine learning algorithm, the future congestion on the network could be predicted up to 75 minutes ahead. This research could also be factored into this project, where the overall delay at an intersection could be tracked and predicted in the future, allowing for early decisions to be made for future events.
Selecting better projects
- A real-time delay model would also translate to an overall reduction to the cost of congestion. An annual Cost of Congestion report defines the best and worst performing intersections on the network, but by transitioning this into a real-time model, the overall operating cost of the network could be significantly reduced by measuring this information each day.
- An accurate model could allow us to track the overall performance of traffic signals, which could determine a cost-benefit before projects go through the Assess and Select phases. This would also allow projects to be tracked during and after completion.
Allow real-time network intelligence
- An accurate model would allow for real-time tracking of performance. This system would take the next step in recommending decisions for signal operations and planned interventions.
This project aims to develop a pilot model that utilises secondary datasets within Main Roads (e.g. signal timing data) to estimate overall delay at each intersection in real-time. An area of the Perth road network will be chosen for the pilot.
The project objectives are to:
- Determine the best analysis and modelling methods to use given available datasets.
- Use existing data to develop a mathematical or data-driven (including machine learning) model for an area of Perth employing agile development principles involving rapid prototyping and fast iteration to arrive at a pilot model from a range of preliminary models
- Validate the accuracy of the models using observed delay time data gathered from traffic surveys or Main Roads’ existing traffic cameras
More from iMOVE