Testing ZF’s Autonomous Ride-Hailing Taxi

We try out a robo-taxi at CES, and learn where the industry really is at on the road to autonomous lifts

As Silicon Valley and the car industry continue to converge, we’re seeing an increasing presence of autonomous cars at the Consumer Electronics Show, including an autonomous robo-taxi that we caught a ride in this year. The ZF Autonomous Ride-Hailing Taxi is a prototype that’s meant to preview the experience of a new form of urban mobility, but that technology will actually be put to use this year in Germany.

While the robo-taxi looks nearly identical to a regular van—the slight difference is in the noticeable sensors that are not on production cars today—the interior is where the experience initially deviates from the one we’re accustomed to. Immediately noticeable is the lack of a steering wheel and foot pedals. There is a joystick that can manually allow control of the vehicle, but ZF’s engineers tell us this is only because they don’t think the public is ready to give up complete control just yet.

We have to be 100 times better than we are with human drivers

ZF’s CEO, Wolf-Henning Scheider, tells us that the psychological barrier to autonomous cars cannot be overcome with safety statistics or numbers, but by people experiencing it for themselves. “Word of mouth will be absolutely key here,” Scheider says, “We have to be 100 times better than we are with human drivers.”

Rob Csongor (vice president of automotive machines at NVIDIA) points out that we’re already using this service in different forms. “You’re already being driven by robo-taxis, and don’t know it. The trams at airports, even elevators, used to be operated by humans. And, early on, people were nervous. But you don’t think about those things now. It will be the same with cars.”

The robo-taxi uses the company’s newest, most advanced processor, the ZF Pro-AI RoboThink, which constantly pulls data from the car’s sensors, including cameras, radar, and LiDAR systems. Indeed it is the computing power that Scheider says has been the biggest breakthrough in the last few years. “The industry has been talking about autonomous cars for twenty years at least, but now we have the computing power to process all of this data.”

The latest ZF ProAI RoboThink developed with NVIDIA has 600 times the TOPS processing power of its first-gen processor, which helps it analyze and react to the constantly changing environment a vehicle encounters. “These are deep neural networks, and we’re essentially using a supercomputer contained within a small space. Autonomous driving is a computing problem, a scaling problem, and the impact on society is immense,” NVIDIA’s Csongor says.

What’s most surprising is how quickly one gets used to being driven by this embedded supercomputer. In the initial moment, there’s a strange novelty; knowing this was a truly new experience. But quickly it resembles an Uber or Lyft ride you might take today. Our prototype robo-taxi ride had a “driver” in the front of the car—though he was doing as much driving as we were in the second row. We can tap buttons on the screen labeled “airport” or “school,” among others. The car then accelerates, brakes, steers, and stops when you reach your destination.

For this test, we’re in a controlled environment, a geo-fenced parameter that engineers can map and program. But that’s exactly how Scheider says autonomous driving will begin. He believes the initial automaker timelines of 2020 and 2021 for level 4/5 driving were overly ambitious. “I won’t give you a date as to when we’ll have this capability, because I can only be wrong. There are legal and legislative issues, the public has to be ready, and there are still several challenges. What we will see however, is autonomous driving in restricted environments, or geo-fenced areas.”

ZF will begin its own autonomous testing in Friedrichshafen, Germany, where the company is headquartered. Using its e.GO people mover, the vehicles will initially have a driver on-board, and the purpose of the first year or two will be to collect data, and hone the system. Eventually the autonomous-capable vehicles will roll out to other cities, as it moves towards autonomy.

Images courtesy of ZF