PALO ALTO, Calif. -- Drifting might seem like an art. Even a skilled race car driver usually will struggle to skid a car around a corner under perfect control.
But it's actually a science: a predictable pattern that a trained driving computer can execute with near-perfect precision.
Last week, by unveiling an autonomous DeLorean that drifts itself in perfect doughnuts as its tires lose traction, Stanford University researchers illustrated the importance of teaching self-driving cars to push the limits of grip and execute the tricky maneuvers mastered by the most skilled human drivers.
Modern cars have electronic stability control to keep them from slipping and sliding. This is useful because most human drivers can't handle a skidding car, says Chris Gerdes, the Stanford engineering professor leading the project.
But a computer can.
"When we no longer have a human driver in the loop, we think that the automated vehicle should be able to harness the full range of vehicle operating capabilities to avoid collisions, even if this means going sideways a bit to stay on the road," Gerdes said.
"We aren't literally envisioning roads full of automated vehicles that can produce clouds of white tire smoke," he added, "though that would be cool."