By Joann Muller

Waymo, which pointedly stopped using the term “self-driving” to describe its technology this year, has released a study intended to prove that its robot drivers are safer than humans.

Why it matters: With about 40,000 Americans dying in vehicle accidents every year, AV operators are trying to convince consumers and regulators that autonomous vehicles would make the roads safer.

What’s happening: Waymo, which operates a limited driverless taxi service near Phoenix, reconstructed 72 fatal accidents that occurred over the past decade in its geo-fenced operating area.

  • It then fed the data from those real-life crashes into its simulation system, and substituted the “Waymo driver” for the human driver.

What we know: Waymo’s autonomous technology avoided or mitigated collisions in almost all cases.

  • When the Waymo driver replaced the “instigator” of the accident — a drunk driver speeding through a red light, for example — the crash was avoided because the robotaxi is engineered to obey the law.
  • When the Waymo driver replaced “the responder” — someone reacting to a bad driver — its perception systems anticipated the situation earlier and responded to avoid it.
  • The few instances where the Waymo driver couldn’t avoid the accident was where it was struck from behind.

Be smart: Waymo’s research is helpful, but not definitive.

  • A study last year by the Insurance Institute for Highway Safety found that autonomous systems — because they can see better than humans — can be expected to prevent about one-third of highway crashes.
  • The other two-thirds are caused by “predicting” or “deciding” errors, and AVs can’t avoid these if they drive too much like humans.

Instead, they must be specifically programmed not just to obey laws but also to prioritize safety over speed and convenience.

  • That could mean intentionally crawling through high pedestrian traffic or in low-visibility conditions.

The bottom line: AVs could be safer — and slower.