While the VW announcement was vague — "The collaboration brings the two companies together to realize self-driving electric vehicles in cities as Mobility-as-a-Service (MaaS) fleets" — Hyundai provided a concrete goal: "a strategic partnership to bring self-driving Hyundai vehicles to market by 2021."
You may not have heard of Aurora, which has been described in some news accounts as "mysterious." But Aurora Innovation has been in business since December 2016, and it is to autonomous technology what the 1927 Yankees are to baseball. The three leaders of the company are Chris Urmson, co-founder and CEO, who had previously been chief technology officer for Alphabet Self-Driving Cars; Sterling Anderson, co-founder and chief product officer, who had directed the development of Tesla Autopilot; and Drew Bagnell, co-founder and chief technical officer, who had been autonomy architect and perception lead at the Uber Advanced Technology Center.
We had the chance to sit down with Chris Urmson after he appeared onstage at a Hyundai press conference. He shared his insights on Aurora's approach to automated driving.
"We think the first place this technology comes to market in in the transportation services or ride-hailing applications, but that's for our partners to decide." (Ride-sharing is a strategy a lot of players in the field are shooting for, as round-the-clock use is one way for paying for what will initially be a technology too costly for private ownership.)
Initial deployment of self-driving cars?
Transporting goods or people?"I personally — and as a company — am more excited initially about moving people around. Urban mobility. That's where you see the largest social impact. And it provides better access to mobility for people."
Can you create a car that doesn't crash?"It is a fundamentally hard problem because other operators on the road can behave erratically at any moment. For example, if you are in a two-lane, opposing-traffic road, if you want to be safe, you don't drive there, ever. The opposing traffic could at some point drift in the lane and at a late enough moment that you would be at an unavoidable collision state."
The importance of vehicle-to-vehicle capability?"I think V2V and V2X [vehicle-to-infrastructure] are interesting sensors. One of the challenges with V2V is that you don't get any real value from it until you get significant deployment. ... I don't believe you need that technology to deploy self-driving vehicles."
Teleoperation?"We believe the actual driving intelligence has to reside on the vehicle. At the end of the day, you can't afford to be in a communication dead spot."
State of tech development?"The software has a lot to be done. It's not complete yet. And like any engineering problem, you can solve it in a variety of ways. There is room for improvement on the sensors. Cameras are in a very good place, and they continue to get better, pushed by the cellphone market. Automotive radar has been well established, but it feels like it is in a local minimum in that there may be some innovative things you could do with the same fundamental technology that would be more relevant for self-driving vehicles. Lidar is another element that has been under-explored, and there is an opportunity to do something exciting there."
The fun of driving?"It's not to say that driving isn't fun and can't be fun. It's that a lot of driving isn't fun. At one point I had my fantasy about driving a nice sports car, but I realized there was really no point because I would just get speeding tickets."