Over my life, I have been involved in various submissions to road authorities about changing traffic conditions in areas the community has identified as dangerous.
The typical response is that the data doesn’t support the claim that a particular location is dangerous. By “data”, that generally means accidents. If there haven’t been enough crashes, injuries or fatalities, there may not be enough evidence to justify changes.
The obvious, and somewhat loaded, response is: “Do we need to wait for someone to die before we fix the road?”
Unfortunately, historically, accident data has been the best information available. Road “black spots” become black spots because enough bad things have already happened there.
The airline industry took a different approach – with outstanding results.
In the mid-1950s, Australian investigators began working on the concept of recording what happened aboard aircraft. Following a fatal Fokker F27 crash in Queensland in 1960, Australia became the first country to make cockpit voice and flight data recorders compulsory on commercial aircraft. The idea quickly spread around the world.
By the 1980s, computers made it practical to analyse enormous numbers of flights, while confidential incident reporting became an important safety tool.
The key change was philosophical. Airlines no longer needed to wait for an aircraft to crash to learn something.
The results have been remarkable. In the 1950s, there was roughly one fatal accident for every 29,000 flights. By the 1970s that had fallen to around one in 165,000 and continued dropping. Today, the figure is around one fatal accident for every two million flights.
Could the car industry learn the same lesson?
It may already be starting.
Modern cars are increasingly connected computers on wheels. That concerns some people – understandably – because these vehicles can collect an extraordinary amount of information. But some of that data may have a very valuable use: identifying tomorrow’s black spots before they earn that terrible title.
Researchers at the University of Technology Sydney have analysed data collected from 700,000 vehicles over 20 months across New South Wales. Rather than looking only at crashes, they examined harsh braking, sudden swerving, high g-force events and other violent vehicle movements.
Think of these as the road equivalent of aviation’s near-miss reports.
Across NSW, the researchers found an average of about 1,000 such incidents every day, sometimes reaching 2,000. When they compared these events with existing accident black spots, they found a strong correlation.
More interesting, though, were locations with lots of near misses that aren’t currently classified as dangerous.
That could give road authorities something they have never really had before: a crystal ball.
Imagine a particular intersection where cars repeatedly brake heavily. Perhaps sight lines are poor. Maybe the traffic-light sequence encourages risky turns. Perhaps a bend consistently causes drivers to swerve.
At the moment, we may discover the problem after several ambulances have visited.
In the future, software could notice thousands of vehicles behaving abnormally at that location and flag it for investigation. Engineers could alter the intersection, reduce the speed limit, improve signage or change traffic-light timing before anyone is seriously injured.
There are obvious cautions. Hard braking doesn’t automatically mean a dangerous road. A child might run onto the street, traffic might suddenly stop or another driver might simply do something silly. Data needs context, and researchers say more work is needed to prove exactly how reliably near misses predict crashes.
This is where connected vehicles could deliver a substantial public benefit.
For decades, road safety has often been about studying yesterday’s tragedy to prevent tomorrow’s. Now we may be able to study today’s near misses instead.
Mathew Dickerson