Tech Talk
Way More Waymo: Why the robotaxi beat Tesla to the gate By Michael E. Duffy L ast month, I described my experience as a passenger in a Waymo “robotaxi.” This month I’d like to delve a bit into the technology behind the
The team at Waymo has its origins in the Stanford University team that participated in the 2005 and 2007 Grand Challenge events organized by the Defense Advanced Research Projects Agency (DARPA) to encourage the development of autonomous vehicles. Waymo began as the Google Self-Driving Car Project in 2009 with the core of that Stanford team. Waymo was spun out of Google in December of 2016. Looking at the history of the project, however, reveals a key element of Waymo’s success: mapping. The original team was hired to assist with Google’s Street View project, where camera vehicles are driven (by humans) along publicly accessible roadways. It’s a feature of today’s Google Maps.
experience. What struck me was how, in many ways, a ride in a Waymo is utterly unremarkable. The car shows up, you get in, it delivers you to your destination. The technology to make that happen is pretty remarkable, though.
As context, it helps to understand a little about how the autonomous driving software that powers Waymo (and others) is categorized. There are six “levels of autonomy” from 0 (no driving automation) to 5 (full driving automation) outlined by SAE International. A level 5 system should be capable of “all driving tasks in all conditions without any human intervention.” At present, there are no level 5 systems. Waymo is widely considered to be a level 4 system, capable of handling most driving tasks in a specific environment. Despite Elon Musk’s claims/hype, Tesla’s Full Self-Driving (FSD) is rated as a level 2 system (partial driving automation). A third company, Zoox (owned by Amazon) is also considered to be a level 4 system, although it has yet to roll out its service to the general public (they are presently testing in Las Vegas). The fundamental difference between Tesla’s tech and that used by Waymo/Zoox comes down to a philosophical choice. Elon Musk believes that since humans manage to drive using only their eyes, Tesla vehicles should be able to get by using only cameras. Waymo and Zoox, on the other hand, use both radar and lidar (LIght Detection And Ranging), in addition to cameras, to navigate their surroundings. One reason for this is cost. Cameras are much less expensive than the sophisticated radar/lidar hardware that gives the Waymo and Zoox vehicles their distinctive sci-fi appearance. Tesla’s original aim for FSD was private vehicles, not robotaxis, and that makes cost a paramount concern. The assumption is that with the appropriate software, a camera-equipped vehicle should be able to drive as well as a human. As yet, Tesla has yet to prove its case. Tesla says it’s going to roll out robotaxi versions of its Model Y in Austin and, as I write this, it appears they have at least one on the streets there. According to Musk, it is not running the same software as in the company’s regular cars. I, for one, am skeptical about any claims from Tesla—it’s been hyping “full self-driving” for years, and has yet to deliver. Waymo, on the other hand, is the acknowledged leader in robotaxis. Why? The most likely reason is simple: it’s been doing this a lot longer.
When riding in the Waymo, I was struck by the accuracy of the maps displayed on screen. And Waymo has spent a lot of time and effort mapping out the cities in which it operates in detail. It only operates within the boundaries of these maps (called “geofencing”). In contrast, Tesla collects map data from the cars it has sold. It’s a lot of data, to be sure, but I suspect that Waymo’s map data is of much higher quality. Again, Waymo’s focus on autonomous vehicles operating within a defined area as its core business has driven it (so to speak) in a different direction than Tesla, which focuses on automobiles which can drive between two arbitrary points. Although I’ve focused on Waymo and Tesla, it’s important to note that China is also working on autonomous vehicles. The Chinese company Baidu also has Level 4 vehicles operating in several Chinese cities. Like Waymo, it operates in specific geographic areas and uses radar and lidar in addition to cameras. Unlike Waymo and Tesla, the software used in Baidu vehicles is open source, which means anyone can use it and work on it. It’s not clear if this will prove to be an advantage, but it does bring outside resources to bear on the problem. Building a vehicle which can drive between arbitrary locations in all conditions is a hard problem. Waymo has constrained the problem by operating only in specific, well-understood areas without much adverse weather. I’m not sure they are trying to solve the general problem—e,g, navigating from L.A. to NYC in the middle of winter. There is a massive opportunity in simply delivering taxi service in urban areas and doing it better than anyone else. g
Michael E. Duffy is a 70-year-old senior software engineer for Electronic Arts. He lives in Sonoma County and has been writing about technology and business for NorthBay biz since 2001.
July 2025
NorthBaybiz 37
Made with FlippingBook Digital Publishing Software