Back
Back
Mercedes-Benz Next: Autonomous driving.
  • Humans and AVs: It’s complicated.

  • Humans and AVs: It’s complicated

    • 15. March 2017
    • Autonomous Driving
    • Illustration: Dave Haenggi
    • Text: Steffan Heuer

    Conference SXSW 2017 in Austin, Texas: Just as software needs to learn how to drive, humans need to learn how to co-exist with increasingly intelligent vehicles.

     

    On the road to a fully autonomous “robot chauffeur”, human drivers and semi-autonomous vehicles have to establish a nuanced relationship to sound out each other’s strengths and weaknesses.

     

    At this year’s South by Southwest (SXSW) conference in Austin, Texas, experts from private industry and government tried to narrow down the issues at stake in a series of panel discussions. The annual tech gathering is putting increased focus on the future of connected mobility this year, featuring more than 30 sessions around the topic.

     

    At one panel, entrepreneurs, engineers and regulators agreed that the interactions between human drivers and the machines they rely on will fundamentally change. In order to reap the benefits of this new technology, including fewer traffic fatalities, they argued that people have to become comfortable with systems they don’t fully understand.

    Higher standards than to ourselves

    “When we relinquish control, we tend to hold it to higher standards than ourselves,” said Nathaniel Beuse, associate administrator of the National Highway Traffic Safety Administration (NHTSA), the US agency charged with setting and enforcing safety standards to reduce accidents and road fatalities.

     

    He asked how much safer autonomous vehicles need to be in order to be considered “safe enough”. Choosing human drivers as the benchmark might be a bad choice because human error led to more than 35,000 deaths in the U.S. in 2016.

     

    The spread of semi-autonomous driving features raises the question how to better train humans in using these new technologies. Unlike airbags or seat belts, most drivers don’t understand the exact workings of advanced driver assistance systems (ADAS) or can’t interpret them, according to Beuse. Neither do they take the time to study manuals. Relying on dealerships to train new owners in how to use semi-autonomous features is also not a practical solution.

    ' Cars have to know us. You can’t put it on humans to truly understand the technology. '

    Axel Nix, Head of ADAS Systems at Harman

    A personal relationship with technology

    To close this knowledge gap, technologists suggest several approaches to train the humans on the road to full autonomy. For one, forming a more personal relationship with technology can help establish trust. “Our interaction with cars will fundamentally change. We will perceive them not as a machine, but like a family member,” said Axel Nix, head of ADAS systems at Harman, traditionally an audio and infotainment brand.

     

    He compared the future experience of interacting with an AV to his children speaking with Amazon’s virtual assistant Alexa at home. “Cars have to know us. You can’t put it on humans to truly understand the technology.” The key question to his mind is one of social acceptance: “When are we ready to tell the driver it’s ok not to pay attention, that the car is safe enough to pick up a book and read?”

    Similar to a computer game

    Chris Augeri, CEO of mobility startup Drive Spotter, proposed a gamification approach to prompt humans to stay within the bounds of new technology. Why not let drivers “unlock” higher levels of ADAS features similar to a new level in a computer game when they behave well, or take away privileges when they don’t pay attention or speed, he asked? Fine-grained data gathering, including video feeds, from vehicles in conjunction with high-definition maps enables this kind of carrot and stick approach.

     

    The computer scientist, who previously worked on drones and remote sensing, also suggested semiautonomous systems should take design cues from consumer apps like Gmail. That would include asking people questions and prompting for basic input or feedback before they use a feature. “Sometimes it’s better not to flash an alert that would confuse you, but to build information into the navigation screen. For example to show the next stop sign because we know you blew through the previous one,” he said.

     

    Catherine McCullough, executive director of the Intelligent Car Coalition, a lobbying group for auto, tech, and telecom companies, raised the point at what speed society is comfortable with testing and rolling out autonomous driving features. “When you say AV, many people think of level 5, a true robot chauffeur. We are not there yet—it will take a while.” But waiting too long to achieve some measure of perfection may also mean to needlessly resign from a much higher level of road safety through robot technology.

    ' When you say AV, many people think of a true robot chauffeur. We are not there yet. '

    Catherine McCullough, Executive Director of the Intelligent Car Coalition (ICC)

    Related.