This article is more than 1 year old

Sergey Brin building humanitarian blimp for lifesaving leisure

You may have to wait a decade or more to join him aloft

Reality check

The rise of drones, technological advances in a variety of fields, and the emergence of on-demand business models – in conjunction with alternative aircraft ownership schemes – have revived hope that personal air transport can work for the salaried class.

But drone companies, the paper notes, are failing, raising concerns about market viability and the projected timelines for the era of on-demand personal aviation. As with self-driving cars, the projected arrival of passenger aircraft capable of electrically powered vertical take-off and landing – what Uber refers to as eVTOL – raises a host of issues that haven't been solved.

Despite the excitement, much of the research and development work remains undone, the paper says, like certifying the safety of probabilistic reasoning algorithms, electric propulsion systems, communication systems and fleet coordination systems.

In a phone interview with The Register, Cummings, associate professor at Duke University and director of Duke's Humans and Autonomy Laboratory, said that the missing pieces – a regulatory framework and the technology necessary to support on-demand air transit – are intertwined.

"For Uber's Elevate vision to happen, you have to have a battery-powered VTOL aircraft," she said. "We don't have the batteries today that can do that job, nor do we have a VTOL system that has ever been certified for passenger carrying."

And without the technology to test, regulators can't formulate meaningful rules.

Cummings expressed confidence that the technical challenges will be solved, but hedged on the ten-year timeline suggested in the paper. "Ten years is academic code for ten, twenty, or thirty years," she said.

Then there's the question of how to design vehicles expected to be largely autonomous. Do you want your self-driving air taxi to have a manual override steering wheel or maybe full flight instrumentation for trained pilots? And when you board your six-year-old and tell Siri to take the child to soccer camp, do you tell her not to touch the red button? Or do you disable it and bet her life on bug-free code?

Cummings suggested that concern about human error, not to mention human malice, will produce designs that deny passengers access to flight controls. And that means we need flawless, tested, predictable AI.

"The core issue, whether it's self-driving cars or autonomous planes, is that we have no idea how to guarantee that probabilistic reasoning algorithms can make acceptable choices in the most safety-critical situations," said Cummings. "Can the plane always guess right if it loses an engine over a congested environment, and not injure anyone?"

When we test airplane engines today, we put them on a block and they can be expected to perform the same way, said Cummings. But driverless cars today perform differently in the same environment. As an example, she said, long shadows might alter how car's cameras perceive the world.

"With AI-driven planes or cars, it's still pretty much the Wild West in terms of how we certify these systems," said Cummings. ®

More about

TIP US OFF

Send us news


Other stories you might like