As it became known from the new report of the NationalThe US Transportation Safety Board, the Tesla Model 3 Sedan, which in March crashed into a truck on the Florida highway and killed its driver, acted with an active autopilot function. This driver was at least the fourth to die in an accident related to the Tesla semi-autonomous driving function. What is striking in the March 1 crash is that the details of the accident are almost identical to the details of the first public report of the deadly autopilot crash in May 2016.
In each case, Tesla on autopilot collides with a truck crossing the path and kills the driver.
The victims of the autopilot Tesla Model 3
CEO Ilon Musk boasted abilitiesself-driving their cars and promised that next year they will become completely autonomous. But critics say the autopilot system — which requires drivers to be alert and ready to take control of the car — does not work well enough to keep drivers in check, and that Tesla shows the system as more capable than it actually is. The new accident casts doubt on how well Tesla responded to this criticism, given its active promotion of more ambitious technologies.
In the March accident, the red Tesla Model 3 was drivingsouth along the right lane of state highway 441 in Delray Beach, about 110 km north of Miami. The truck left the private road on the right side of the highway, intending to turn left and head north. The report says that the truck slowed down, crossing the lane, "blocking the Tesla road."
The Autopilot system on Model 3 turned on in about10 seconds before the accident and the car did not see the driver’s hands on the steering wheel for eight seconds immediately preceding the impact. The representative of Tesla confirms this. The car crashed into a trailer at a speed of 110 km / h (with a speed limit of 90 km / h), not trying to carry out any maneuvers to dodge. The car drove under the truck, tearing off the roof and killing the driver Jeremy Beren Banner, age 50 years. Truck driver was not injured.
The report does not report how farTesla, when the truck went on the road, and there are no weather conditions that could affect the ability of the car to slow down. It is therefore unclear whether a driver who does not use the autopilot could stop safely. Rough mathematics suggests that it would take the Model 3 driver a few seconds to assess the situation and avoid an accident. At a given speed, the car travels 30 meters per second. Braking distance Model 3 at a speed of 100 km / h is about 40 meters. The car could stop for 75 meters. We add 1.5 seconds to the driver’s braking actions and it would seem that three seconds would be enough.
On May 7, 2016, 40-year-old Josh Brown diedvery similar circumstances. His Model S worked on autopilot when he was driving north on Highway 27A in northern Florida. Towards left the truck. Tesla did not stop the movement and crashed at a speed of 125 km / h (with a limit of 110 km / h), passed under the truck, tore down the roof and killed Brown. Then the car drove another 100 meters, hit the pole quite hard and stopped 15 meters later. Truck driver was not injured.
Then it became obvious that the truck driverwas supposed to give Tesla, and Brown was inattentive. But some of the blame rests with Tesla, which allowed the driver to overly rely on the automation system for a long period.
After Brown’s death, Tesla shortened the timefor which the driver may not touch the steering wheel before the system issues audible and visual warnings. Replaced several iterations of hardware and software. But the circumstances of the accident indicate that when driving at high speed, the system is still unable to detect some stationary objects or objects moving perpendicularly to the car. Probably for the same reason, at least three Tesla crashed into a standing fire truck in 2018 (without serious injury).
Model 3, which was in the last accident, wasequipped with cameras that should, theoretically, detect a truck crossing the path of the car. Tesla did not respond to a request from Wired about how the system uses radar and cameras or what steps were taken to avoid such accidents.
At the same time, Tesla is optimistic aboutMask-based computer vision capabilities that should allow “full independent driving,” according to Mask, without the need for human observation. Last month, Andrei Karpati, head of Tesla Autopulot's technical department, said Tesla believes it can use machine learning methods and cameras to make its cars first-class drivers. Mask went even further and said: "I quite confidently predict the appearance of Tesla autonomous robotaksi next year."
What do you think, will Tesla realize its promise? Tell us your opinion in our chat room in Telegram.