Move to Driverless Cars Slower Than Anticipated, Experts Agree

Experts see a shift in focus for wireless automotive technologies from a focus on driverless vehicles to features that make conventional cars safer to drive, experts said Wednesday during a Keysight webinar. Experts agreed getting more autonomous vehicles on the road is taking longer than expected a few years ago.

In 2015, a lot of people thought that by now everyone would be riding around in autonomous vehicles or “robotaxis,” said Sam Abuelsamid, principal analyst at Guidehouse Insights. “Clearly that’s not the case,” he said. Automated vehicles have been “a lot more difficult than anybody anticipated,” he said: “It’s a lot harder to prove that these systems are actually safer than human drivers.” Driverless cars are seen as one of the biggest use cases for wireless as carrier move beyond 5G.

A lot of the technology that has been developed is being used in wireless driver-assistance systems rather than driverless cars, Abuelsamid said.

An estimated $160 billion has been invested in the technologies needed for driverless vehicles, said Vijitha Chekuri, Amazon Web Services global head-automotive compute services. The rule is that developing 90% of a technology takes 10% of the time, she said. Getting more automated vehicles on the road is “taking longer and longer” and “the timelines are shifting,” she said.

Chekuri agreed a lot of the technology is “shifting down” to making driving safer, including lane keeping, lane assistance, traffic-jam assistance and similar features, she said. Some technology once available only in high-end cars is already available in less expensive vehicles, she said.

Watching autonomous vehicles get to scale is like “watching water boil,” said Aaron Newman, Keysight business development manager-autonomous vehicle technologies. “You know it’s going to happen, but when, exactly, you’re not sure,” he said.

Driverless vehicles require more than AI and sensors, Newman said. The required infrastructure and ecosystem are advancing, and policymakers are developing the regulations, he said. “All of those things are happening as we watch,” he said. Driverless taxis are “growing in what they can and how much they’re allowed to do,” he said. They’re still only in limited areas but the footprint is growing, he said. “At some point, the water will start boiling,” he said.

Companies will have to form partnerships for driverless vehicles to be successful, Chekuri said. “If you look at what it takes to develop highly automated autonomous vehicles, there are many, many elements to it,” she said. She cited the need for data-management, simulation and machine-learning tools, hardware and infrastructure.

“No one company has it all, end to end, perfectly,” Chekuri said. Data management is an issue, with vehicles collecting terabytes of data every day, she said. More sophisticated vehicles could have 50 sensors operating “and you’re talking about 80 TBs of data per day,” she said.

Eventually, automated vehicles will be able to react faster than human drivers “and probably more accurately,” Newman said. The human driver is the backup when power steering or another system fails. With a driverless vehicle, “you’ve got to have multiple layers that act as a backup to keep the vehicle going until you can at least come to a safe stop,” he said.

A lot of the engineers working on driverless vehicles are from other parts of the tech industry, and had to adjust to taking a more careful approach, Abuelsamid said. The philosophy has been “move fast and break things, and ship a minimum viable product,” which is a “very low bar,” If a photo-sharing app crashes, “the consequences are pretty trivial,” he said. That’s not the case for a 5,000-pound car, moving on the road surrounded by other vehicles and pedestrians, he said. “The consequences of making a mistake are much higher -- people die,” he said.