With 1,753 people killed and 25,638 seriously injured on NSW roads between 2016 and 2020, improving safety is a crucial goal for the government agency Transport for NSW (TfNSW). And data is their secret weapon.
TfNSW already analyses safety data from the Australian Road Assessment Program (AusRAP) for 18,000km of the most heavily used parts of the road network and uses it to develop ‘star’ risk ratings for roads and target improvements.
This is being constantly improved by incorporating a wide range of internationally-verified data sources – including vehicle navigation systems – and machine-learning techniques.
That’s where a team of scientists led by Dr Simona Mihaita from the UTS Data Science Institute came in as part of a TfNSW research project through the iMove Cooperative Research Centre.
“The current road network that has been assessed for its road star rating has undergone mostly a manual process in the past and represents only 13% of the total NSW road network," Dr Mihaita says.
"In reality, there are multiple road types from motorways, freeways, highways, arterial roads and sub-arterial roads that require an automatic road star rating in the future.”
TfNSW brought the team on board to work with large data providers such as TomTom and geospatial providers like Anditi.
"We came together to apply international standards and find new ways to use artificial intelligence and machine learning to reduce the costs and improve the identification of attributes that will improve road safety outcomes,” she says.
The result was world-leading research capturing attributes to include in the road ‘star’ risk ratings.
It increased the number of attributes that could be automatically analysed to international standards from 13 to 34.
Some data points analysed as part of the project included speed limits, operating speeds, road curvature, grade, skid resistance and number of lanes.
“We looked at a large Multinet-R database from TomTom road layout and overlaid it against existing AusRAP data, making easier the attribute extraction,” Dr Mihaita says.
“This meant we could automatically consolidate road safety data from pre-existing sources at 10 per cent of the cost to manually do custom survey precision data. We also looked at the scalability of such an approach with great promises in other states and countries worldwide.”
These are advanced steps that the team from TfNSW can use to improve the effectiveness and reduce the costs of their risk management approaches and infrastructure investments.
Peter Dunphy, Head of Transport Safety at Transport for NSW said the road toll is not just a number, it is real people.
“To ensure road safety is addressed proactively across NSW, there's a need for more efficient and cost-effective data to drive road safety decision making on State, regional and local government roads," he says.
“The iMOVE research project has proved that new artificial intelligence and machine-learning based data collection and feature extraction methods, delivered as part of the global AiRAP data partnerships, will improve the accuracy, reliability and scalability of capturing AusRAP data for the State’s network and beyond.”
Other Australian road safety authorities and international organisations can also leverage the findings to improve road safety data models further.
The next stage in the research will be to improve the TfNSW data management by investigating AiRAP-compliant data associated with road user flow (pedestrians, cyclist flows) and speed, to pilot ‘star’ ratings for smaller roads and to further improve the model.
The project AiRAP automation for Australian road safety was managed through the iMOVE CRC by lead agency TfNSW in partnership with UTS, the International Road Assessment Programme (iRAP) TomTom and the geospatial data company Anditi.
