NTSB: Autopilot Design Flaw, Inattentive Driver Led to Tesla-Firetruck

NTSB: Autopilot Design Flaw, Inattentive Driver Led to Tesla-Firetruck

The NTSB said a combination of a flaw in Tesla’s highly regarded (certainly by Tesla) Autopilot system, plus driver inattention, caused a Tesla Model S to slam into an on-call firetruck parked on a California freeway in January 2018. No one was injured, but the 2014 Tesla Model S P85 needed more than Dent Wizard repairs.

The National Transportation Safety Board said:

[T]he probable cause for the crash was the Tesla driver’s lack of response to the fire truck parked in his lane, due to his inattention and overreliance on the car’s advanced driver assistance system; the Tesla’s Autopilot design which permitted the driver to disengage from the driving task; and the driver’s use of the system in ways inconsistent with guidance and warnings from Tesla.

An NTSB timeline of the January 2018 crash in which a Tesla Model S rear-ended Engine 42 of the Culver City, CA, fire department. Orange segments show when Autopilot was engaged and the driver was hands-off the wheel. The crash is the right side of the timeline.
An NTSB timeline of the January 2018 crash in which a Tesla Model S rear-ended Engine 42 of the Culver City, CA, fire department. Orange segments show when Autopilot was engaged and the driver was hands-off the wheel. The crash is the right side of the timeline.

The NTSB recreated the 66-minute trip that ended in the crash (graphic above), with the help of Tesla’s onboard data recorder and various sensors. It found the driver was hands-off (on and off) for 12 of the 13 minutes leading up to the crash. The driver is supposed to always have hands at least lightly on the wheel to allow safe Level 2 self-driving. The latter means the car stays centered in its lane (if there are no sharp curves, this is almost never a problem on interstates) and maintains a set speed and/or paces the car in front. Most drivers of Level 2 cars soon learn they can get away with giving the steering a little jiggle every 10-15 seconds and in some cases, even less often.

Here’s what the NTSB said in this week’s preliminary finding (NTSB accident ID HWY18FH004):

The NTSB’s investigation revealed the crash trip lasted about 66 minutes, covering about 30 miles, with the “Autopilot” system engaged for a total of 29 minutes, 4 seconds. Hands were detected on the Tesla’s steering wheel for only 78 seconds of that 29-minute, 4-second period. For most of the time the system was engaged, it did not detect driver -applied steering wheel torque (hands on the steering wheel). The “Autopilot” system issued several hands-off alerts during the last 13 minutes, 48 seconds prior to the crash and was engaged continuously during those nearly 14 final minutes of the crash trip. In the last 3 minutes, 41 seconds before the crash the system did not detect driver-applied steering wheel torque.

During most of the driver’s operation with the “Autopilot” engaged, the system detected and followed a lead vehicle, one that was ahead of the Tesla. In the 15 seconds prior to the crash the system detected and followed two different lead vehicles. Data show that 3 to 4 seconds before the crash, the lead vehicle changed lanes to the right, a movement commonly referred to as a “cut-out scenario” in testing and research. When the Traffic-Aware Cruise Control no longer detected a lead vehicle, the system accelerated the Tesla from about 21 mph toward the preset cruise speed of 80 mph, which had been set by the driver about 5 minutes before the crash. The “Autopilot” system detected a stationary object in the Tesla’s path about 0.49 seconds before the crash and the forward collision warning activated, displaying a visual warning and sounding an auditory warning. By the moment of impact, the Tesla had accelerated to 30.9 mph.

In simple terms, Autopilot finally recognized the firetruck when the two vehicles were 40-45 feet apart (basic math: a vehicle at 30 mph covers 44 feet per second), about two car lengths. The 47-year-old male driver told the NTSB he bought the Tesla in part because he could use the HOV lane to get to work in LA from his home in Woodland Hills (median home price: $815,000) without taking on a passenger. He was, however, accompanied by a cup of coffee and a bagel, but can’t remember if he was drinking coffee or eating the bagel when the car struck the firetruck and the airbag went off. (Memo to NTSB: Check the shirt for spatter marks.) He bought the Model S used, did not real the manuals, but did have the car safety-checked and says the Tesla staff explained the workings of Autopilot.

As a result of the accident, the report says, the NTSB went to the makers of Level 2-autonomy cars in the US and asked what they were doing to develop apps to “more effectively sense the driver’s level of engagement and alert the driver when the engagement is lacking while automated vehicle control systems are in use.” It went to VW Group (which includes Audi), BMW, Nissan, Mercedes-Benz, Tesla, and Volvo. According to the NTSB, “All manufacturers except Tesla have responded to the NTSB.”

So, while the driver was not paying fullest attention, the blame for the crash, according to the NTSB, also goes to the “Tesla Autopilot’s design … which permitted the driver to disengage from the driving task … and the driver’s use of the system in ways inconsistent with guidance and warnings from the manufacturer.”

Continue reading

Western Digital Changes Its Reported Drive Speeds to Reflect Reality
Western Digital Changes Its Reported Drive Speeds to Reflect Reality

Western Digital has launched new WD Red Plus models to correct previous communicated inaccuracies regarding the spindle speeds on its 8TB-14TB products in this family.

Intel Announces New Optane Drives, 3D NAND SSDs
Intel Announces New Optane Drives, 3D NAND SSDs

Intel has announced a number of Optane and NAND storage devices coming to market in 2021, with a 3rd-generation Optane product on the roadmap thereafter.

New Study Proposes Warp Drive That Might Actually Work
New Study Proposes Warp Drive That Might Actually Work

Physicists have speculated on the possibility of a real warp drive for years, but a new paper lays out a vision for a warp drive that might actually work. We still don't know how to build it, but at least we know why we can't build it.

Perseverance Takes Its First Drive on Mars
Perseverance Takes Its First Drive on Mars

Perseverance finally took its first drive on Mars late last week, covering a total of 21.3 feet (6.5 meters).