The company explained in a blog post over the weekend that engineers had asked employees and their little trick-or-treating ghosts and goblins to trek around some parked autonomous cars. That gave the cars' sensors and software extra practice time recognizing kids in different shapes and sizes -- and even costumes and masks.
"Halloween's a great time to get some extra learning done," wrote the team from Google's Self-Driving Car Project. "We teach our cars to drive more cautiously around children. When our sensors detect children -- costumed or not -- in the vicinity, our software understands that they may behave differently."
Engineers want the cars to recognize children and be aware that they are more apt than adults to dart into the road, be obscured by parked cars or run down a sidewalk, chasing a ball.
Google continues to put miles on its driverless car effort, giving the cars time to maneuver autonomously on the highways and city streets .
Safety's important: Even more of Google's self-driving cars are on the road today than there were this past spring.
By late September, Google had licenses for 73 autonomous cars in its fleet. That's more than triple the 23 licenses it had last May.
The company also has been talking with executives at car companies in Detroit, looking for a partner to build its autonomous cars one day.
Googlers may be working on the software and artificial intelligence to run driverless cars, but the company doesn't necessarily want to get into the automobile manufacturing business. For that, Google would like to team up with an experienced manufacturer.
However, it looks like Google is going to have some competition.
Early in September, it was reported that Toyota is teaming up with Stanford University and MIT to work on the artificial intelligence needed to make the auto manufacturer's cars more autonomous.
Toyota is investing up to $50 million in the project over the course of five years.