Uber has been testing semi-autonomous vehicles in a handful of cities for a few months now, but a new report suggests that its tests aren’t exactly going smooth. While issues are to be expected with any sort of testing phase, the data we’re seeing today suggests that Uber is having a hard time making significant progress. This, in turn, could prevent Uber from rolling out a fleet of autonomous vehicles within the timeframe it has envisioned.
If you’ve read coverage about Uber’s self-driving tests in the past, then you know that Uber’s fleet of self-driving vehicles – currently deployed in Pittburgh and Phoenix – aren’t truly autonomous. They still require a person behind the wheel to take over when the need arises, and internal Uber documents obtained by Recode suggest that’s a pretty frequent occurrence.
For the week of March 8, these documents show that Uber drivers needed to take over once every 0.8 miles driven. The records go back to the first week of February, and while there is some variance each week, the number stays at or below the one mile mark for the measured period. There is no distinction between the reasons that the drivers needed to take over, as this simply covers driver intervention for any reason at all.
Another graph shows the average miles a car will go before the driver needs to make a “critical intervention.” As you can imagine, critical inventions are quite a bit more serious than taking over to correct a missed turn or perform a lane change – they’re made to avoid hitting a pedestrian or avoid collisions that would cause significant property damage.
The worst week Uber had for critical interventions was the week of February 8, where cars only made it an average of 50 miles before the driver had perform one. The best, on the other hand, was the most recent week recorded – during the week of March 8, cars in Uber’s fleet drove an average of nearly 200 miles between these serious events.
That is definitely encouraging news, and it stands in rather stark contrast to the data that shows regular interventions aren’t getting much better. Uber’s cars are driving more and more miles each week too, so while this data may not be showing much in the way of improvement, it’s also good that we aren’t seeing much of a decline either.
Still, this probably isn’t what Uber had in mind when it set out to test self-driving technology in real world scenarios. Uber, of course, didn’t want to comment when asked, and that probably isn’t going to change. Perhaps we can count on another leak to give us an update on the situation a few months from now? In any case, be sure to check out the source link below to read Recode’s full report.