Queensland scientists are set to go on a road trip with CHAD, but they will not be letting him drive.
Subscribe now for unlimited access.
or signup to continue reading
Yet.
The Cooperative and Highly Automated Driving pilot program, a partnership between the Queensland Department of Transport and the Queensland University of Technology, is testing how well artificial intelligence handles Australian roads.
Professor Michael Milford, from QUT’s Australian Centre for Robotic Vision, said while the technology around driverless cars had improved, it needed to be tweaked for Australian conditions.
“We’ve already found that a lot of that technology actually doesn’t work that well ‘out of the box’ in Australia, so one of the main jobs of the team is to improve the technology,” Professor Milford said.
“By the end of the project we’ll have a much better idea about how well these autonomous vehicle technologies will adapt to Australian road infrastructure.”
The initial trial will see engineers drive CHAD on 1200 kilometres of roads across Queensland, through a range of conditions - night and day, rural and urban, wet and dry.
The test car has been fitted with the full array of sensors and cameras which would be on a driverless car, but instead of letting CHAD drive, the researchers will collect the data of how the system reacts to the conditions.
Queensland Transport Minister Mark Bailey said the information gathered would help inform how the government developed policy around driverless cars.
“What we are seeing here is the future unfolding on Queensland roads. This data will inform the Cooperative and Highly Automated Vehicles of the future,” Mr Bailey said.
“The research no doubt will be shared as part of the process of moving towards a safer road system for everybody.”
The researchers will be looking at how the AI interacts with four specific road categories - lane markings, traffic lights, street signs and GPS “black spots” like low coverage areas or tunnels.
Past studies have shown AI driver software has trouble with roads without lane markings, which can be a problem in rural and regional parts of the state.
Early testing by QUT researchers has also found paint spills and other road debris can be interpreted by the software as lane markings, causing the car to drive erratically.
The researchers found the software would often simply take its hands off the virtual wheel when faced with some of the challenges Australian drivers deal with every day, disengaging autonomous control and leaving it up to the human driver to get them out of situations.
The 1200-kilometre road trip is expected to kick off later this month, and will take about three months.
