Google's Self-driving Cars Still Haven't Learned How Junctions Work

By Gary Cutlack on at

Google's self-driving car ambitions -- and therefore the self-driving car plans of the rest of the world -- look as far away as ever right now, as team members working on the tech giant's autonomous vehicles are slowly revealing the mountains of work needed to get stupid computer eyes to work out if something's a flashing red light or just the enraged face of a van driver shouting at a cyclist.

The Register has analysed the public output of Google hardware engineer Daniel Rosenband, who seems to be autonomously hitting his head against many desks when trying to solve some key problems. Problems including simple junctions, of which Rosenband says: "At four-way stops, oftentimes cars arrive sorta at the same time and it's a coin flip for who goes first. We have to make it comfortable for the person in the car; you don’t want the vehicle to inch forward and then slam the brakes, and you also want to be courteous to other drivers."

Human techniques at such junctions include making eye contact with the enemy drivers to see if they've spotted you or not -- something self-driving cars aren't able to do. So the junctions of the future may be clogged by self-driving cars, all too scared to make the first move.

And that's just junctions. The problems faced in getting computers to work out what traffic lights mean are outlined by Google's tech team here [PDF], where they conclude that, well, it's hard. They'll never manage a US supermarket's mass of mini roundabouts. [The Register]


Want more updates from Gizmodo UK? Make sure to check out our @GizmodoUK Twitter feed, and our Facebook page.