‘Full Self-Driving’ clips show Teslas trying to drive on train tracks and smashing into a bike lane
Tesla’s self-driving technology is slammed as ‘flawed’ by experts after car owners’ videos showed vehicles steering into oncoming truck, failing to stop for pedestrians and hitting roadside obstacles
In footage recorded earlier this month by YouTube user AI Addict, a Tesla Model 3 with Full Self-Driving can be seen smashing into a bike lane bollard in San JoseDuring the same drive, the feature attempts to drive down some light-rail tracks, mistaking it for a roadAnother video posted by the YouTuber in December shows it failing to stop for a pedestrian in a crosswalk In yet another jarring clip, a Tesla vehicle using the technology fails to detect on oncoming truck as it prepares to make a turn, forcing the driver to interveneThe footage highlights some of the flaws with the still largely untested software, according to experts
<!–
<!–
<!–<!–
<!–
(function (src, d, tag){
var s = d.createElement(tag), prev = d.getElementsByTagName(tag)[0];
s.src = src;
prev.parentNode.insertBefore(s, prev);
}(“https://www.dailymail.co.uk/static/gunther/1.17.0/async_bundle–.js”, document, “script”));
<!–
DM.loadCSS(“https://www.dailymail.co.uk/static/gunther/gunther-2159/video_bundle–.css”);
<!–
Tesla‘s beta stage Full Self-Driving software does not appear to be ready for the mass market after users posted multiple videos in recent months showing harrowing near-misses.
In footage recorded earlier this month by YouTube user AI Addict, a Tesla Model 3 sporting the technology – which is still being tested but is available to selected Tesla owners in its incomplete ‘beta’ stage – can be seen smashing into a bike lane bollard at 11 mph on streets of San Jose. During the same drive, the feature attempts to drive down some light-rail tracks, mistaking it for a road.
AI Addict also posted a video in December of his car failing to stop for a pedestrian in a crosswalk.
Meanwhile, YouTuber Dirty Tesla shared a terrifying clip in November of his Full Self-Driving software preparing to cross traffic into a left turn – ignoring a fast oncoming truck. The driver was forced to intervene, grabbing the wheel to avoid collision.
Tesla CEO Elon Musk claimed only last month that the FSD Beta system had never had an accident, but the new footage offers proof that this is no longer the case.
The jarring videos highlight some of the flaws with the still largely untested software, which uses a combination of cameras and ultrasonic sensors to stitch together a view of the world so cars can drive autonomously, experts say, highlighting the unpolished feature’s failure to detect common road obstacles and pedestrians.
Scroll down for video
In footage recorded earlier this month by YouTube user AI Addict, a Tesla Model 3 sporting Full-Self Driving- which is still being tested but is available to selected Tesla owners in its incomplete ‘beta’ stage – can be seen smashing into a bike lane bollard at 11 mph in San Jose
‘The video [footage] shows different scenarios where the automated driving system was not able to detect and/or cope with relevant features of its Operational Design Domain,’ Nicola Croce told The Washington Post Thursday, referring to the conditions under which the Full Self-Driving system is expected to safely operate.
Croce currently works as a technical program manager at Deepen AI, a company that helps carmakers employ driver-assistance and autonomous-driving technologies in their vehicles. Tesla is not one of its clients.
The slip-ups, Croce told the paper, reveal an inherent weakness in the software that resulted in it not meeting basic safety requirements it was designed to adhere to – a sentiment that was echoed by other autonomous-driving experts who viewed the videos.
Speaking on the clip of the car crashing into the bike lane divider in San Jose, Brad Templeton, a seasoned self-driving-car developer who worked on Google’s self-driving car, said the cause of the collision was due to the software being incomplete.
‘The bollard issue is both mapping and perception,’ Templeton told The Post. ‘As permanent bollards rather than temporary cones, they should be on a map,’
That way, he said, ‘the car would know that nobody ever drives through these.’
‘As to why the perception missed them until too late, this is an issue with computer vision. Perhaps it never got trained on these unusually shaped and [colored] bollards,’ he added.
As for the near-miss with the pedestrian, who in the clip was standing just off the sidewalk at the start of a crosswalk when the car fails to stop to the driver’s horror, experts blamed the lapse on the technology’s failure to recognize pedestrian walk signs, or foresee the fact that a stopped pedestrian might venture off a sidewalk.
‘It’s unclear whether the car reacted or not to [the pedestrian’s] presence, but clearly the driver is shaken,’ said Andrew Maynard, a professor at Arizona State University. Maynard serves as the director of its Risk Innovation Lab, where he works on technologies similar to Full Self-Driving.
Hod Finkelstein, chief research and development officer for AEye, a company that sells light detection and ranging (lidar) technology to carmakers seeking to implement smart cameras into their vehicles, theorized that the incident occurred because cameras alone are insufficient when it comes to detecting pedestrian intent.
Finkelstein says this is because cameras are not capable of measuring the distance of distant objects, and can be blinded by glares originating from headlights or the sun.
The jarring videos highlight some of the flaws with the still largely untested software, experts say, highlighting the unpolished feature’s failure to detect road obstacles and pedestrians
A Tesla Inc Model 3 electric vehicle, like the one seen in one of the videos, is seen here displayed inside a showroom in Tokyo, Japan
Elon Musk in August admitted that Tesla’s new self-driving software is ‘not great’ but the firm is trying to fix it
Tesla boss Musk famously sought to use only cameras and ditch sensors in his vehicles’ less sophisticated Autopilot feature, which was panned by experts after at least 10 were killed in eight accidents in which Tesla’s Autopilot was engaged since its release in 2016, reports from the National Highway Traffic Safety Agency reveal.
Last week, Tesla had to recall nearly 54,000 vehicles equipped with its ‘Full Self-Driving’ software, after it allowed vehicles to run through stop signs at low speeds, without coming to a complete halt.
The company also had to recall over 800,000 vehicles because seat belt reminder chimes may not sound when the vehicles are started and the driver isn’t buckled up.
All were to be fixed with online software updates, where possible.
Safety advocates and automated vehicle experts say Tesla is pushing the boundaries of safety to see what it can get away with, but now NHTSA is pushing back.
In November, NHTSA said it was looking into a complaint from a California Tesla driver that the ‘Full Self-Driving’ software caused a crash.
The driver complained to the agency that a Model Y went into the wrong lane and was hit by another vehicle.
The SUV gave the driver an alert halfway through the turn, and the driver tried to turn the wheel to avoid other traffic, according to the complaint. But the car took control and ‘forced itself into the incorrect lane,’ the driver reported.
NHTSA also is investigating why Teslas using the company’s less-sophisticated ‘Autopilot’ partially automated driver-assist system have repeatedly crashed into emergency vehicles parked on roadways.
The agency opened the investigation in August 2021, citing 12 crashes in which Teslas on Autopilot hit parked police and fire vehicles. In the crashes under investigation, at least 17 people were hurt and one was killed.
Last week Tesla said in its earnings release that ‘Full Self-Driving’ software is now being tested by owners in nearly 60,000 vehicles in the US. It was only about 2,000 in the third quarter. The software, which costs $12,000, will accelerate Tesla’s profitability, the company said.
Tuesday’s recall is the 15th done by Tesla since January 2021, according to NHTSA records, with almost all of the more than one million vehicles sold in the US requiring an update due to a recall.