Forget highways and interstates, the real test of any driver (parallel parking aside, of course) is the ability to navigate congested city streets without the loss of life, limb or hubcap. Even Google is aware of the challenge, and it's putting its self-driving car to the urban test.
An official Google blog post Monday revealed the next phase of the self-driving car project has begun on the streets of Mountain View, California, where over the last year the revamped vehicle logged thousands of miles around the Internet giant's hometown.
Chris Urmson, director of the self-driving car project, said in the blog post:
We've improved our software so it can detect hundreds of distinct objects simultaneously — pedestrians, buses, a stop sign held up by a crossing guard, or a cyclist making gestures that indicate a possible turn. A self-driving vehicle can pay attention to all of these things in a way that a human physically can't — and it never gets tired or distracted.
The software puts to work a set of complex algorithms to perform the city street feat, analyzing a combination of data gleaned from the notorious spinning bucket on the rooftop, 64 lasers scanning 360 degrees, a camera, and GPS map data. An image is then rendered for the vehicle that separates obstacles into colors:
Google said the vehicle sees the chaos and randomness of the city street as predictable, with software models built to react to both likely and unlikely scenarios.
Google's ability to master self-driving algorithms could put the pressure on the likes of Ford and GM to also come up with solutions to make roads safer by way of automated driving features. During anat CES 2014, Ford CTO Paul Mascarenas noted the development process for such features is far from rapid, and is rather "a continual progression of more advanced functionality."
Google also admits its work is far from complete:
We still have lots of problems to solve, including teaching the car to drive more streets in Mountain View before we tackle another town, but thousands of situations on city streets that would have stumped us two years ago can now be navigated autonomously. Our vehicles have now logged nearly 700,000 autonomous miles, and with every passing mile we’re growing more optimistic that we’re heading toward an achievable goal — a vehicle that operates fully without human intervention.
Beyond mastering the algorithms, there is still work to be done on the software modeling front. A software modeling platform is required for autonomous city vehicles operating in non-simpliﬁed urban trafﬁc conditions, and perhaps an element that will speed the autonomous vehicle's time to market.