Waymo’s rapid expansion is exposing a persistent challenge for autonomous vehicles: handling unpredictable situations that fall outside their training. During San Francisco’s July 4 fireworks celebration, driverless Waymo vehicles blocked intersections, worsened congestion, and left some motorists stranded for hours. Several cars ran out of battery power, while at least two drove over exploding fireworks, tells The New York Times (full article available to subscribers).
Such incidents are known in the autonomous vehicle industry as edge cases. They are becoming more visible as Waymo, owned by Alphabet, expands its fleet. The company now operates nearly 4,000 autonomous vehicles across 15 metropolitan areas and provides about 500,000 paid rides each week. Since December, however, Waymo has issued three federal recalls related to software problems.
Critics argue that situations described as edge cases are often routine features of urban transportation. Large events, emergency scenes, construction zones, and unusual traffic patterns regularly confront human drivers. Researchers also warn that autonomous systems may struggle to recognize when they have encountered unfamiliar conditions and need assistance.
One incident involved a Waymo entering a closed highway construction lane, swerving through traffic cones, and passing workers while police officers attempted to stop it. Waymo later recalled software on some vehicles and temporarily suspended freeway routes while developing a fix.
Waymo maintains that its technology is improving as it encounters more real-world situations. The company says its safety record shows 94% fewer serious-injury crashes than human drivers.
Regulators and San Francisco officials are nevertheless seeking stronger safeguards. Federal authorities are developing autonomous vehicle safety standards, while city leaders want clearer operating rules during emergencies and major public events.
The debate highlights a central challenge for autonomous driving. Expanding robotaxi services requires more than performing well under ordinary conditions. These systems must also recognize and respond safely when city streets become unpredictable.
