Monitoring where cars frequently swerve and brake may allow authorities to identify potentially dangerous locations and take action to make them safer before crashes happen.
“We strongly believe that the presence of near-miss data is evidence that a crash will occur in that area,” says at the University of Technology Sydney in Australia.
She and her colleagues have just analysed severe braking and swerving data collected from 700,000 vehicles between April 2020 and December 2021 throughout the state of New South Wales.
Advertisement
The information was recorded by data platform from onboard accelerometers uploading real-time incidents of “harsh vehicle movements, violent manoeuvres, near misses, collisions or high g-force incidents”.
Mihaita says the accumulation of near misses indicates risky driving behaviour that just “keeps building up and up until you’ll have to deploy rescue teams, or you will have collisions or crashes happening in that area”.
The research was focused on south-east Australia, but the team thinks identifying danger spots before crashes occur should be a global priority.
The research revealed that across New South Wales, there were an average of 1000 near-miss incidents every day, with a maximum of 2000.
The worst areas for close calls were within the Greater Sydney area, with the city itself recording well over 40,000 near-miss incidents during the research period. Only in the Parramatta local government area, to the west of Sydney, were figures higher: here, over 45,000 incidents were recorded.
When these incidents were overlaid on known accident black spots, there was a strong correlation but, says Mihaita, many locations where near misses are happening on a regular basis aren’t yet recognised by authorities as dangerous.
“Many areas that are considered safe and ignored by authorities are not that safe at all,” she says.
The current approach is to first “wait for people to end up in crashes or die and then take some measures”.
Instead, says Mihaita, authorities could proactively intervene in areas where dangerous driving and near misses are known to occur by reducing speed limits, improving infrastructure or increasing monitoring and law enforcement.
“I’m very passionate about people not dying on our roads,” she says. “We have a responsibility to carry everyone safely from their home, to schools, to childcare, to their work, and back as safely as we can. I think in Australia we have the means to make that happen.”
Unpublished data collected by Compass IoT shows that the number of near misses across Australia is continuing to grow dramatically – reaching nearly 200,000 in 2025. In the same year, Australian road deaths reached 1337 – the highest in 12 years.
at Griffith University in Brisbane says many decisions regarding road safety interventions are made using retrospective data – after crashes have occurred.
“More proactive approaches are useful because they can prevent deaths and serious injuries from occurring in the first place,” Bates says.
at RMIT University in Melbourne says he is cautious about whether data on near misses can be used to predict actual crashes and that more validation work is needed to confirm the link.
However, he says the data is a “potential early-warning signal for road-safety problems, complementing traditional crash data”.
These events, he says, are useful indicators of driving conditions or behaviour, but need to be interpreted in context.
“A harsh manoeuvre doesn’t necessarily mean dangerous driving or a near crash,” says Shiwakoti.
“A driver may brake hard because a pedestrian has stepped onto the road, traffic has suddenly stopped, another driver has behaved unpredictably or the road environment requires a sudden manoeuvre.”
arXiv
