City Science, in partnership with the University of Exeter, was awarded funding from Innovate UK to develop an end-to-end process for the automated production of transport models from open data sources, incorporating automated calibration of assignments to TAG standards.
Overview
In many cases transport modelling is unnecessarily the delay or in conflict with decision making, including in regards to transport decarbonisation. There is an increasing urgency to not only make transport modelling faster, but also more customised to address pressing issues such as climate change. To combat this, we identified various aspects of conventional model builds that can be automated and developed bespoke modelling tools to assist with a range of transport challenges. Automating these processes means that the models and data behind them are highly transparent and can be efficiently integrated into a range of workstreams. This cuts out the repetitive tasks between projects and allows more time for detailed analysis and innovative solutions.
Scope
Working with our long-term research partners at the University of Exeter, City Science was awarded funding from Innovate UK to develop a process for the automation of the building of transport models. This extensive research programme focused on the use of game-changing technologies within the transport modelling world. This included not only the automation of the model build, but also the automation of assignment calibration using novel optimization and computational techniques.
City Science Response
Regardless of study area, building transport models requires a number of standard base tasks; automating these, whilst still recognizing the unique features of each area, results in consistent and efficient model builds and allows more time for the detailed analysis required. Our award-winning team of software developers and transport technologists developed an automated model building process, underpinned by end-to-end data pipelines. The model build continuously checks dependencies within the project to ensure that a new model can be run as a single, seamless process for any location in the UK.
The project included the development of:
- Flexible processes to define areas of network detail to align to TAG.
- Standardised data pipelines for network building including speed, volume, and capacity attributes derived from open data sources, such as OpenStreetMap.
- Processes to automate structures required for data processing, such as zoning systems.
- Generation of base-year synthetic origin-destination demand matrices, through implementing a furnessing and gravity model process (drawing on National Travel Survey (NTS) and National Trip End Model (NTEM) data).
- Processes for automated end-to-end running and analysis of the models, including visualizing differences between model runs.
In addition, we have developed new methods to automate the calibration of large-scale transport models to effectively calibrate speed-flow curve parameters. This was a significant achievement which demonstrates that, with sufficiently robust input data, the assignment calibration process can be fully automated, vastly simplifying the production of models that comply with TAG standards. Alongside the automated calibration, we have developed a calibration and validation assistant tool allowing for clear and detailed visual comparisons of assignment outputs with observed data, including journey times, screenline flows, and convergence. These tools allow for a streamlined calibration process.
Outcome
City Science has automated the processes involved in a conventional transport model build, using extensive software code to enable us to develop models for any location in the UK based on OpenStreetMap data. We have tested innovative solutions with respect to calibration, establishing a leading model for the automation of the calibration process.
These tools have been extensively tested and implemented across a range of projects. They continue to be used to enhance the quality, detail, and delivery of our work, ensuring that our transport models comply with TAG guidelines.