Robotaxi operators learn from stumbles
Add Axios as your preferred source to
see more of our stories on Google.
Illustration: Aïda Amer/Axios
Robotaxis, like any new driver, often make mistakes — as more and more of them get deployed in U.S. cities, the tech companies behind them are learning important lessons from things they could have done better.
Why it matters: Every incident — every dumb or potentially dangerous error — is being scrutinized as they work to earn public trust.
In one infamous incident, many of Waymo's robotaxis stopped operating and blocked traffic during a chaotic power outage in San Francisco that knocked out traffic signals and cellular networks.
- Waymos are trained to call for remote guidance when they encounter a traffic light that's not working.
- But when an entire section of the city went dark, remote operators were quickly overwhelmed with requests, explained Allison Drutchas, a Waymo attorney, at NHTSA's AV Safety Forum.
- The usual safety redundancy was actually counterproductive in that situation, she said.
"What we learned from that is to have the capability to sort of identify areas or times when that extra level of conservatism is no longer appropriate, and the car should sort of trust its opinion and proceed, even if it hasn't gotten a response from remote assistance," she said.
- That means treating the intersection like a four-way stop, and yielding appropriately to other traffic, she explained.
In another incident, some Zoox robotaxis got stuck in an area blocked by police activity in Las Vegas, a problem that took about 10 minutes to resolve even with the help of remote assistants, said John Maddox, senior director of safety strategy and operations.
- "We're treating it as a huge learning opportunity," he said.
- "If we don't learn from those events, we will never, ever get better. We'll never be able to scale fully, so it really is incredibly important to learn from your operations."
The bottom line: Waymo's co-CEO, Tekedra Mawakana describes the challenge as "advancing safely."
- "How do you advance a technology that is so nascent, while earning public trust, while demonstrating that the benefits are worth the reasonable risk exposure that comes from a technology that cannot be completed in a lab?"
