Tests of truly autonomous self-driving vehicles have been uneven at best, but with each scraped test hubcap, the concept draws closer to reality. Major automakers are in, with Toyota backing autonomous driving startup Pony.ai and Fiat Chrysler striking a deal with AutoX, the Hong Kong-based firm that has developed an artificial intelligence (AI) platform for driverless vehicles.
The AutoX-Fiat Chrysler pact to get driverless robotaxis up and running in China this year is ambitious, as there is still much to solve. Likening it to the “moonshot” efforts of the 1960s, Dr. Jewel Li, chief operating officer at AutoX, recently told PYMNTS’ Karen Webster that “autonomous driving has so many components … from the cloud, from simulation to software, to the AI, to decision-making, to hardware. It’s a very complicated system engineering problem.”
But backed by very serious investors — like China’s Dongfeng Motor Group and eCommerce titan Alibaba, among others — the AutoX platform is solving tech orchestration issues required for successful autonomous journeys. Their progress is impressive enough to have earned one of only two autonomous vehicle operating permits granted by the state of California to date.
In April, AutoX and Alibaba’s AutoNavi app partnered on a ride-hailing service in Shanghai.
Obtaining these golden credentials is made possible by AutoX’s groundbreaking work in the Chinese city of Shenzhen, where its autonomous vehicles have been operating. Politely calling the local driving style there “aggressive,” Dr. Li noted that “the quality of the data, the density of the data, is significantly higher than when you test” in less crowded environments. “That helped us move a lot faster, and broke a lot of our assumptions as well.”
The Trust Factor
The psychology of driverless cars, from both the passenger and pedestrian perspectives, is a fascinating part of AutoX’s efforts to blend into urban environments, like any other kind of vehicle. Li said the company has recorded the reactions of passengers in autonomous cars, who tend to be nervous at first. A screen in the vehicle displays route selections and other “decisions” being made by the AutoX AI as it navigates unpredictable city streets.
“[Passengers] continuously [check the screen] to make sure the system is doing everything great,” Li said. “But after five minutes, nobody’s looking at [the screen] anymore. After 20 minutes, we’ve seen people falling asleep in the car. People give this technology a lot of trust.”
With the explosion in urban mobility in recent years, particularly with bicycles and scooters, autonomous vehicles must maneuver streets that are more crowded than ever before.
Recognizing objects like other cars is only the first step in the AI’s “education”, Li told Webster. “The harder challenges are, for example, pedestrians. It’s not just ‘let’s recognize there is a human,’ but it’s having to look in such detail that we … know the intention of the pedestrian” before the car decides its next move, he explained.
Subtle details, like pedestrian eye movement, signal an intention to cross the street. The AutoX AI calculates these and other cues as it plots each route, correcting as a human driver would.
Autonomous Use Cases, Delivered
If robotaxis are the sizzle, the other things autonomous vehicles can do are the steak. Use cases arising around smarter autonomous vehicle AI include trucking, all manner of logistics and a host of other applications where driverless vehicles would actually be an advantage.
But Li hastens to note that “we’re creating a ‘driver’ that can drive different vehicles. That’s the goal. It can do other things as well. We can use a similar type of vehicle for ‘last-mile’ delivery, and we can also drive a light truck,” which she said shows strong B2B potential.
“When you create an AI driver, the data and the effort is more focused on how to deal with all the users on the road. It matters less about what vehicle you’re driving,” Li noted. As time goes by, it also matters less who owns the self-driving car, as Li pointed to the growing popularity of fractional and subscription car ownership models.
Speaking about the AutoX U.S. pilot program, Li noted that as in Shenzhen, density and depth of data is critical. “We’re focused on being fully autonomous, and we will start that in San Jose, where we have the highest amount of data in the U.S. to ensure the highest safety.”
Whatever the use case — ridesharing, freight hauling or food delivery — the real breakout moment for autonomous vehicles is coming. It just has an obstacle course to get through first.
“The core dependency [we face] right now is still the autonomous driving technology stack itself,” Li said. “And we need to make sure that it’s safe to handle a scaled-up amount [of vehicles on the platform] and also [a geographic] area.”