Uber’s self-driving cars are marvels of modern technology. Unfortunately, they also don’t seem to know what a one-way street looks like.
A video posted to Facebook shows one of Uber’s autonomous Ford Fusions undergoing testing in Pittsburgh. By “undergoing testing,” I really mean making a turn down a one-way road, before the human realised the mistake and made a large and embarrassing U-turn in the intersection.
The incident seems to have ended without any accident or problems, but it does highlight how far autonomous cars have left to go before they can fully match human drivers. Uber is testing four self-driving Ford Fusions in Pittsburgh, with bigger plans to roll out a fleet of 100 self-driving Volvo SUVs later this year. All self-driving cars still have two humans to help ensure safety, one designated driver and another engineer in the passenger seat.
From the looks of the video, any competent human driver wouldn’t have made a turn there. Even if you didn’t see the road sign, the double line of cars facing one way is a dead giveaway that it’s a one-way street. Human drivers with any experience driving in a city would recognise that without even thinking about it, but that’s a far more nuanced thing for a car to try and see.
Uber’s self-driving technology appears to rely on a LIDAR unit mounted on the roof, and presumably a set of cameras. In some way, the sensors are superior to a human: they’re constantly monitoring the road, and are much better at sensing humans hiding between cars or keeping tabs on ten other cars at once.
But the sensors (and the computers interpreting their signals) are still far worse at understanding nuance than a human driver. In the short term, that probably means we’re going to see a lot more embarrassing videos like the one above. In the long term, it’s going to be a real problem for rolling out fully autonomous vehicles driving around wet, rainy cities at night.