Teaching machines how to drive
Just a couple of years ago, there was much talk of how autonomous vehicles were only a whisker away and were poised to enter the mainstream.
But since Queensland scientists completed a 1,200km road trip across the south of the state, reality seems to be setting in.
“We have more realistic expectations now, in that nobody actually knows for sure when they will be commonplace on the roads,” said Michael Milford, a professor of electrical engineering and the researcher who led the QUT test drive.
“Over the last two years we have seen a fairly significant shift in public commentary. People were predicting with certainty that there would be so many autonomous vehicles by now.
“But they now have a more cautious outlook whereby they don’t think they will be widespread tomorrow and it may actually take many, many years—apart from a few isolated ones like Tesla,” he added.
Milford’s project took an electric Renault fitted with hi-tech sensors onto the roads to analyse the response of artificial intelligence to a wide variety of driving situations. It set out to determine if AI could perform the same as a human driver in all conditions.
In particular it looked at how a car could use infrastructure such as lane markings and road signs to help it drive well. This was particularly apt in Australian conditions.
“There aren’t many other places where you will see a ‘kangaroos crossing’ sign, so it throws up its own challenges.
As the three-month project draws to a conclusion, the results are currently under wraps, and it is not known when, if ever, they will be revealed to the public. But they do highlight the growing focus that the tech industry is placing on the things autonomous technology cannot do, as opposed to all the optimistic and often wild predictions of what they can do.
“Like all tech, some stuff will come a lot sooner, some stuff will be much slower than you expect or may never even occur,” said Milford.
“People are a lot more savvy now about the key barriers to autonomous vehicles becoming widespread. Things like those one-in-a-million incidents like when a guy in a chicken suit runs across the road. Those are the critical problems that are still to be solved.”