A new branch of real-time analytics that predicts when and where a road incident is likely to happen has captured the interest of Main Roads, as well as freight and transport services across Australia.
Rationalising data from cameras, road sensors, weather conditions and historical data, scientists can predict possible future incidents, traffic jams and how long it will take for the road to clear.
“Algorithms are used to predict the likelihood of an incident occurring and the duration of the accident,” said Michael Borck from Curtin University Department of Computing.
“It might be bumper-to-bumper traffic, a set of failed traffic lights or a truck that has lost its load. Based on the predictions, the transport department can decide whether to re-route traffic.”
According to the RAC, nose-to-tail accidents are Perth’s most common type of collision, and most often due to driver impatience and inattention on congested roads.
Mr Borck said Main Roads are planning to use predictive modelling to improve traffic flow and reduce stop-start conditions on Perth’s northbound Kwinana Freeway.
“Main Roads are looking at having a variable speed on the freeway to increase the overall flow. At certain times of day, the emergency lane becomes a legal lane, so in peak hour traffic they can open up that and reduce the speed to get greater flow going through.”
Sensors installed on the onramps and overpasses will monitor usage and wear and tear, and assist with the maintenance of the road, alongside vehicle detectors and CCTV, and new emergency stopping bays at the side of the freeway.
In the event of an impending bottleneck or incident, traffic light timings may be altered on arterial roads and vehicles directed to use alternative routes.
While live predictions can avert impending disaster, Mr Borck said they also expose accident black spots.
“Where it may mean changing the phasing of the traffic lights in the short term, if it’s becoming a pattern and a bad spot, then the transport department will try and make longer term solutions and create alternative access ways.”
In Sydney, data from airline arrival times is integrated in the algorithms and used by freight services to optimise business operations.
“Being aware of delays and unplanned incidents means they can re-route their drivers and improve their services,” said Mr Borck. “Long-term, they can use the predictions to establish new routes and reduce transit times.”
But while intelligent traffic systems could merge information from numerous devices, assessing the impact of doing so is problematic.
“It is hard to tell how accurate the predictions are, or whether the accident would have happened if prevention measures had not been taken,” said Mr Borck.
“And there’s no guarantee just throwing more data at something will solve the problem because some of it may be correlated or make the same prediction – and which one is better? There’s a large research area on that.”
Mr Borck, who teaches predictive analytics at Curtin University, said people, not artificial intelligence should be making the final decisions.
“While the system may suggest the optimal changes, it may not apply common sense. It’s easy to put a lot of faith in artificial intelligence, but it really comes down to how you’ve built it. There is always going to be a human in the middle.”
