They can tear around a prepared autocross course, tackle urban streets and even drive on the freeway — but don’t think that Google self-driving cars are ready for prime time yet.
That’s the message from various academics in the field of transportation technologies and from the director of Google’s automated driving team himself.
As MIT Technology Review details, Google’s self-driving cars have been used to great effect by the software giant as the poster children of near-future transportation, and they’re now a common site in the states and cities which have granted Google permission to test the vehicles out on the public highway.
But while it’s easy to get carried away in the dream of the future, self-driving technology is harder than it looks.
Essentially, teaching the car to speed up and slow down, steer itself and use its radar sensors to brake in an emergency is relatively easy. Teaching it to react dynamically to road signs and changing road conditions is not.
What’s more, preparing a road for the kind of self-driving technology being developed by Google requires hours of preparation and drive-by passes in a specially-designed sensor vehicle.
Think how long it’s taken Google to map the world using its Google Maps software. Now imagine that every single street sign, stop light and lane choice from Google’s extensive Maps and Google Earth database needs to be correctly interpreted and double-checked for autonomous driving purposes.
If a traffic light, stop sign or intersection is changed, so too must the software’s database. At the moment, humans can deal with out-of-date GPS and map data by making choices about where they should go next. At the moment, Google’s self-driving cars can’t do that.
Chris Urmson, director of Google’s self-driving car team, says that Google has already thought of how to tackle the issue of outdated map data and moving signs, but declined to tell MIT Technology Review exactly how it would be carried out.
As Tesla CEO Elon Musk has pointed out in the past and Nissan has more recently admitted by dialling back its own expectations of autonomous driving technology, building a self-driving car capable of driving 98 percent of all roads is reasonably easy — but it’s the final few percent that prove difficult.
In Google’s case, that smaller percentile includes things like teaching the cars to differentiate the difference between a rock and a crumpled piece of paper, or the difference between a pothole and an uncovered manhole.
As Urmson admits, Google’s self-driving cars can’t do either yet. Nor have they driven on snow or in rain, two road situations they’ll need to master before taking over from us behind the wheel.
Think back to your own first few years behind the wheel however, and we’re sure you’ll remember the challenges of encountering a strange new road layout, night-time obstacle or stray road debris for the first time. Years later, we’re sure you still sometimes find yourself debating what to do next in an emergency too — we know we do.
Critics of the way Google has displayed its self-driving program say that Google is lulling the public into a false sense of security by implying that self-driving cars are just around the corner. Google’s own team say that they’re still confident that the majority of challenges will be met and solved in short order.
Self-driving director Chris Urmson says he has a personal target: finishing development of Google’s self driving car in time for his son’s sixteenth birthday. That’s just five years away.
Will he succeed? Will Google really start selling and making self-driving cars in five short years? Or is driving just too difficult for a computer to master in such a short period of time, despite massive advances in processor power?
Leave your thoughts in the Comments below.
You can also support us directly as a monthly supporting member by visiting Patreon.com.