An early report issued by the National Transportation Safety Board (NTSB) Thursday reveals that the Tesla Model X involved in a fatal crash on U.S. Highway 101 in Mountain View earlier this year sped up just before it collided with a traffic barrier.
The preliminary four-page report sheds new light on the March 23 accident in Mountain View, which is currently being investigated by the federal safety agency. The 2017 Tesla Model X struck the barrier between southbound Highway 101 and the state Highway 85 carpool flyover at high speed, smashing the front end of the vehicle. The driver, San Mateo resident and Apple engineer Walter Huang, 38, was extracted from the vehicle and transported to a nearby hospital with major injuries, where he later died.
It was later revealed by Tesla officials that the Model X had the company’s Autopilot feature activated in the moments prior to the crash, which the NTSB report describes as advanced driver assistance including cruise control and autosteer “lane-keeping assistance.” The vehicle’s cruise control speed was set to 75 miles per hour, which ratchets downward based on a set “following distance” between the Tesla and slower-moving vehicles in front of it.
A detailed account of the Tesla’s vehicle logs found that the driver had used the Autopilot system on four separate occasions during the 32-minute trip, including 19 continuous minutes prior to the collision. In the final minute leading up to the crash, the driver had his hands on the steering wheel for 34 seconds, but not during the six seconds prior to the collision.
Within the span of eight seconds, the Model X was following a “lead” vehicle and maintaining a 65 mile-per-hour speed, began a left-steering motion to while following the lead vehicle, and began accelerating when it was no longer following a vehicle. The vehicle sped up from 62 miles per hour to 70.8 miles per hour before striking the crash attenuator in the highway median, with “no pre-crash braking or evasive steering movement” detected, according to the report.
The NTSB report emphasized that the findings are preliminary in nature and not a final determination on the “probable cause” of the crash.
“The NTSB continues to work with the California Highway Patrol and the California Department of Transportation to collect and analyze data, including all pertinent information relating to the vehicle operations and roadway configuration,” according to the report.
In a blog post shortly after the crash, Tesla officials wrote that the damage to the Model X was so severe because the crash attenuator had been damaged in a prior crash, providing less of a cushion between the vehicle and the highway median. The NTSB report confirms that the attenuator had been damaged in a crash on March 12, 11 days prior to the fatal accident, and a large portion of the safety cushion had been sheared off.
Shortly after the collision, the vehicle’s 400-volt lithium-ion battery was damaged and caught fire, prompting the Mountain View Fire Department to blast the vehicle’s interior and exposed portion of the battery with 200 gallons of water and foam, according to the NTSB report. The department also received support from Tesla engineers before determining the battery had cooled enough to transport to a San Mateo impound lot.
The report also recounts how, five days later on March 28, the vehicle’s high-voltage battery reignited, which was extinguished by the San Mateo Fire Department.
Tesla originally partnered with the NTSB to investigate the crash, but was dropped from the investigation in April after NTSB officials said the company released “incomplete information” that was neither vetted or confirmed by the NTSB. The information, released by Tesla in a series of blog posts and statements to the media, implied that the driver was at fault for the fatal crash.
“Such releases of incomplete information often lead to speculation and incorrect assumptions about the probable cause of a crash, which does a disservice to the investigative process and the traveling public,” according to the April 12 statement.
Last month, Bloomberg reported that Tesla CEO Elon Musk had hung up on NTSB Chairman Robert Sumwalt during an exchange about the company’s statements and the decision to kick the company off the investigation. Tesla officials claimed that it was the company that withdrew from the investigation, rather than getting ousted, prior to the April 12 announcement by the NTSB.
Not sure what the information on the battery (fire department and reignited) has to do with the crash.
With all these crash reports, I am shocked that Tesla has not disabled the “autopilot” feature on all its cars. Drivers should drive the old fashioned way until Tesla fixes its safety.
The term auto pilot is what is at fault. It is not auto pilot, but a tool to steering like cruise control is to speed. Both can be used incorrectly.
The fact that he sped up just before the crash makes me question his mindset. I am beginning to look at this a different way.
Did the driver do something at that moment to cause the car to speed up or did the “autopilot” autonomously accelerate the car? My reading of the article is the latter.
The article also says that the autopilot speed was set to 75mph, which is much higher than the legal speed limit and also likely much higher than what is safe for the road conditions.
“The fact that he sped up just before the crash makes me question his mindset. I am beginning to look at this a different way.”
Maybe not. This scenario seems consistent with the published facts.
Tesla’s “autopilot” is a superhyped combo of what other brands offer as automatic lane assist and adaptive cruise control. This particular vehicle’s cruise control had been set to a max of 75 MPH, but the Tesla’s computer sets its actual speed to trail the car in front of it. When the car in front of it went away, the Tesla perceived a clear road ahead and tried to accelerate to 75. The barrier (which its obstacle sensors apparently didn’t register) stopped it at just over 70.
The real issue is the Tesla’s inability to distinguish the inter-lane space containing the barrier from a legitimate traffic lane. Thus it’s “autopilot” happily guided it along in a phantom “lane” which led it directly to the barrier.
The Tesla may have initially followed a car which wandered into that inter-lane space, which is called a “gore”. Its driver recognized his/her error and exited the gore. The Tesla’s robot then saw an open lane in front of it and obediently sped up.
CalTrans needs to reconsider its lane markings so they don’t confuse “smart” cars which really aren’t all that bright. In the meantime, Tesla drivers should closely monitor their “autopilots.”
Please keep your hands on the wheel.
Why are these death traps allowed to operate on the roads of California? The software clearly has bugs. If the crash didn’t kill him, the resulting inferno caused by the battery fire might have.
Luckily no one else was seriously injured. All you need is an innocent party in a non-Tesla vehicle to be killed in an accident caused by a Tesla.
It sounds like to me the sensors in these cars and the processing power is being scrimped on to save money when the described situation is a question of mis-perception on the part of the car. The strategy of designing for good-enough, and then fixing bugs as they happen to approach acceptability has long been used in the software industry, but it is completely unacceptable for something like driverless cars.
In the book “Altered Genes, Twisted Truth” the GMO industry evangelists repeatedly expressed, not only from industry representatives but from government officials, was: “If the American people want progress, they are going to have to be the guinea pigs.”
No one voted on this! If it was expressed openly and actually covered on the nightly news and in the papers or wherever what passes for news outlets in this “news-desert” of America, where the public could hear it, they would be aghast and up in arms to be treated so objectively.
The idea is that basically is that the opinion falsely expressed by industry is that if Americans want to stay at the top some of them have to take the risk and die for the rest of us while we debug these technologies … i.e. find out what we are doing because we do not know. The idea is false because as usually all these costs and risks are born by uninformed Americans who have not voted on or been informed about any of it – so as to save money for these corporations and insulate them from the risks everyone else takes.
This is just pure “socialism for corporations and the rich elites” …. otherwise known as fascism, in terms of assuring profits for the folks at the top, and something far worse than merely being ignored for regular people. Are lives are bring pried into and recorded for reasons and policies we know nothing about, and our health and very existences threatened as we are used like cattle to keep these oligarchs in charge and pushing us about.
Maybe Palo Alto has been mostly purged of all the normal regular working people, or at least Palo Altans think it has, and that might explain some of the support for policies that concern “other people”, but those outside of Palo Alto ought to hear and keep abreast of what is happening here.
Not too long ago “the news” reported on a woman who was killed by a driverless car which didn’t “see” her as she walked her bike across the road. It is clearly a bug in the software that the car didn’t stop and avoid hitting her.
The incident was captured on the vehicle’s video camera. As best I can tell, she was not in a crosswalk. In California she would have no legal recourse.
It appears that this unfortunate driver regularly used “autopilot” without his hands on the wheel during his commute. And he was aware that “autopilot” steering into the barrier to 85 in the carpool lane. “autopilot” sped up toward 75mph after it stopped following the car in front of it but then crashed directly into this barrier. It failed.