Earlier today, on Taiwan’s National Highway 1 near the Zhongshan High Chiayi Water Section, a Tesla Model 3 crashed right smack into the roof of an overturned truck trailer. It appears that the wreck was caused by the Tesla’s Autopilot system not detecting the large rectangular object right in front of it, in broad daylight and clear weather. We have seen Autopilot failures like this before, especially ones involving large, immobile objects.
There’s video of the wreck, and you can see the Tesla drives right into the truck, with only what looks like a solitary attempt at braking just before impact:
????????????
Tesla Model 3 plows info overturned truck on highway. I’m sure the driver was paying complete attention to the road and wasn’t relying on autopilot because he was told the car could drive itself….$TSLAQ pic.twitter.com/cHjueqH0j4— hand washing rooster ???? (@jsin86524368) June 1, 2020
Here’s the opposite angle:
— hand washing rooster ???? (@jsin86524368) June 1, 2020
…and here’s a compiled video from a Taiwanese news site:
For any human driver paying even the slightest bit of attention, this accident is almost an impossibility, assuming the driver had the gift of sight and functional brakes.
Tesla’s Autopilot system primarily uses cameras for its Autopilot system, and previous wrecks have suggested that situations like this, a light-coloured large immobile object on the road on a bright day can be hard for the system to distinguish.
In general, immobile objects are challenging for emergency automatic braking systems and autonomous systems, as if you use radar emitters to trigger braking for immobile objects, cars tend to have far too many false positives and unintended stops than is safe or desirable.
LiDAR systems, which Elon Musk has famously rejected for Teslas, could likely provide better range information for large, immobile objects like this than Tesla’s all-camera setup does.
Also, the understanding that the Tesla, like every car available on the market today, is not capable of fully autonomous driving would help a lot as well, since the driver should be paying attention even with Autopilot active, because this is precisely the kind of situation where an alert — or, hell, even semi-alert — human would have seen that something is wrong and taken steps to, you know, not crash right into that truck.
Luckily, nobody was seriously hurt, but the Model 3 was trashed and covered with what one source called “sauce”:
All that and not a single airbag in the car appears to have gone off. That should be the bigger story here pic.twitter.com/ssi9339vOO
— hand washing rooster ???? (@jsin86524368) June 1, 2020
News reports from Taiwanese outlets, clumsily translated by machine, do seem to suggest that the driver, a 53-year-old man named Huang, had Autopilot activated:
“The Fourth Highway Police Brigade said that driving Tesla was a 53-year-old man named Huang, who claimed to have turned on the vehicle assist system at the time. It was thought that the vehicle would detect an obstacle and slow down or stop, but the car still moved at a fixed speed, so when the brakes were to be applied at the last moment, it would be too late to cause a disaster.”
Tesla’s marketing for Autopilot is certainly confusing, possibly even deceptive regarding the level of capability it has; that said, there are plenty of warnings for drivers to remain alert, so there’s potentially plenty of blame to go around here.
At this moment, though, the takeaway is that no Tesla or any currently-available car is fully autonomous, Level 2 autonomy is not enough for a driver to ignore what’s happening, and the machines are not yet at a point where they can reliably avoid certain obstacles that humans would have no problem with, like this giant, overturned truck.
I’ve reached out to Tesla for comment, but considering we’ve heard fuck-all from them for months and months when we reach out, I’m not going to hold my breath for a response.