Weekly Brief: Waymo AV Data Key to Building Consumer Trust

Waymo claimed last week that its autonomous vehicles are outperforming human drivers.

In a report it compiled, between January 2019 and September 2020, the company’s fleet of AVs logged 6.1 million miles in Phoenix, Arizona. Sixty-five thousand of those miles were without a safety driver behind the wheel. Waymo says that its fleet was not responsible for a single accident in that entire time. There were 18 minor accidents in which AVs were involved. Most of those resulted when an AV was rear-ended by another human driven vehicle.

Only in one incident did one of its vehicle strike and initiate contact with another car’s bumper. That was the freak episode that you may recall from this past February, when a disgruntled former employee hunted down a Waymo vehicle in the middle of the night, swerved in front of it and slammed on his brakes to force an accident. The man, Raymond Tang, was arrested.

To be sure, the company’s AVs are not faultless and the report is careful to point that out. Some of the rear-ending episodes happened because the AVs drive more cautiously than human drivers and, thus, act in ways that human drivers don’t anticipate. These included turning more slowly across intersections and slowing down more gradually than humans expect. In addition, the company says that its safety drivers had to intervene 29 times to avoid potential collisions over the 21 months. Waymo labels these as “simulated contact events.” All of these interventions, the company says, were to avoid accidents that would have been caused at the fault of a human driver. It also claims that not once did its AVs drive off the road.

All in all [if it’s to be believed – Ed] it’s a thorough and impressive track record but the company stopped short of hailing these results as superhuman. The farthest it went was to say was: “Waymo has found the collision frequencies observed in this data to compare favorably to analogous frequencies observed in naturalistic driving studies.” Translation: We’re doing well here. This is especially true since the company claims to have reported every incident that happened, down to minor episodes like a pedestrian walking into a parked AV.

Compare that to the stuff that human drivers regularly don’t report, like swerving off a road half asleep at night or tapping bumpers in a parking lot and mutually agreeing to not get insurance involved. Waymo says it reported everything and its vehicles still outperformed human drivers in terms of accidents and incidents reported.

The takeaways here are twofold: first, the operator could be ready to start scaling more quickly into suburban or relaxed urban environments. Its data suggests that the greatest risk that AVs pose is minor fender benders because human drivers expect them to be more reckless and less law-abiding than they are in practice. Too bad. It’s still a net positive, with the potential for fewer people to drive drunk, tired or distracted. Around 0.02% of the world population dies every year in car accidents and some 40,000 Americans. Waymo’s data and real-world experience suggest that it can help.

Take-away number two: this is how the AV industry needs to build confidence in self-driving cars. Methodically, carefully, with utmost transparency and rigorous data. One could only wish that Elon Musk would take notes. Instead, he has recklessly turned a group of consumer drivers into beta testers of Tesla’s Full Self-Driving platform, as I detailed last week. Thankfully there haven’t been any crashes yet but the technology has been performing erratically so far. You can find dozens of consumer testimonials on Google. In one, the driver exclaims: “I swear I’m not drunk you guys, I’m not drunk, it’s my car.” It seems only a matter of time before something bad happens, given that most consumer drivers are not professionally trained and Tesla drivers in particular have a penchant for showboating and fiddling with technology instead of paying attention to the road.

This is not what the AV industry needs. Just last week, yet another report found wide-spread distrust of autonomous tech among American drivers. This should come as no surprise. Handing over control to a computer is an inherently anxiety-inducing decision, especially in these early days when the practice is uncommon. It’s like flying on a commercial airplane must have been back in the 1950s. If it wants to overcome that anxiety and build confidence, the AV industry should follow Waymo’s lead, not Tesla’s.

Leave a comment

Your email address will not be published. Required fields are marked *