An Australian robotics researcher has hit out at autonomous vehicle makers, saying current solutions are not fit for Australian roads, with more collaboration needed to counter the issue.
Professor Michael Milford, acting director of the Queensland University of Technology (QUT) Center for Robotics, warned that current out-of-the-box European mapping solutions “don't recognize unique Australian signs or infrastructure” and require more customization.
Driving in the land down under
Australia as a nation relies on its roads. The rail network is not suited for passenger and freight travel to some areas of the country, and its large area yet low population density means getting from one place to another is mostly dependent on road travel.
Milford had researched mapping for autonomous cars and used artificial intelligence to see how autonomous cars could handle Australian roads.
He found that keeping maps up to date is proving difficult for autonomous cars all over the world, and while widespread adoption of self-driving vehicles is years away, the work needs to be done now to improve the service.
“We need to plan and design technology that is fit for purpose from the very beginning, not shoehorn it in at the very end when we realize the tech doesn't do what it's meant to do,” Milford said.
The academic called for collaboration between map creators, location services that let vehicles know where they are on a map, and governments for infrastructure updates and privacy regulation, describing such an effort as “the ideal solution.”
“Current maps do not have all the information necessary to be full HD maps or links to information about infrastructure changes,” Milford said.
“Unless a car knows explicitly about environmental changes like road works, for example, positioning systems will find it hard to work well.”
The researcher said government notifications around these types of events are going to be very important but public sector agencies must have meaningful involvement or oversight because of the significant data and privacy implications of such maps.
Professor Milford added that positioning, which formed a core part of the technology offering from autonomous vehicle companies, may also need improvement to provide more accurate services.
“Current positioning systems work well most of the time, but there are failure points like heavy rain and tunnels where the technology is arguably not yet reliable enough.”
“There’d be nothing worse than a car thinking it's in one location, but actually being in another and erroneously referencing the wrong section of the HD map as a result of that positioning error.”
QUT, which specializes in robotic and autonomous vehicle positioning research, is working with the Australian government on the future of HD maps, investigating the ideal models for government-industry collaboration.
“If we started a staged approach toward this collaborative model now, within two years we would have a working prototype for how information from private map providers, the government, and possibly from vehicles on the road, could be shared between all of those key stakeholders to ensure maps are as accurate and up to date as possible,” Milford said.