Featured Tesla Autopilot recall probed by safety regulator following new crashes

Teaching AI basic social skills and how to read body language is the weakest link in self-driving cars. A computer can't look at a pedestrian and instantly decide if they're about to walk in front of the vehicle of it they're pausing just slightly to wait to walk after the vehicle has past. Also the hand gesture of a pedestrian waiving a car past them, versus a pedestrian putting a hand up to demand they stop, isn't something AI does very well.

Also, as the self-driving tech executives in San Francisco learned when their entire business fell apart, you might want to have a camera under the car to make sure that if there's a person trapped under their you don't want to pull over to a safer spot. The fact that they tried to hide this mistake from investigators is one of the greatest dumb moves of all time, but tech executives have so much money, they aren't used to paying attentiont to what they do wrong.

 

ho hum yawn ... so subsequent to Tesla proactively continuing to improve their system .... NHTSA after the fact & maybe starting to feel irrelevant, step up to pretend they're help is helpful
kind of, 'day late & dollar short' of them? Bureaucracy in action

.

 

why is musk so resistant to driver attentiveness?

 

Because all those online videos of people sleeping while the car was driving them helped to sell alot of Teslas

 

The first cited reference in the article, "(NHTSA) said on Friday it was opening an investigation following Tesla’s "

Ok, so I opened the link and found:

• Title and Subtitle, "Final Report of New Car Assessment Program FMVSS 305 Indicant Testing of a 2017 Tesla Model X 75D 5-Door MPV NHTSA No. O20175000"
• Report Date June 7, 2017

I have a problem trusting a story that in the second paragraph cites an NHTSA report nearly 7 years old. But the first paragraph states:

another setback with its Autopilot software​

On Monday, April 21, I drove 147 miles using Full Self Driving and returned Wednesday, April 23. I bought Autopilot in March 26, 2019 and have used it after three weeks of testing ever since. My car has 130,000 mi with the vast majority on Autopilot and a growing number of Full Self Driving(FSD) miles.

My lying eyes and five years driving experience with AutoPilot and FSD says "setback" is inaccurate. Worse, the second paragraph does not cite the original NHTSA notice. Two paragraphs and two 'falsehoods', I would suggest caution.

In 2009-2010, I researched NHTSA investigation processes during the "Bell the Hybrid" nonsense. Today, our hybrids and EVs make 'space ship' like synthetic noises because they were falsely considered too quiet for our streets. Meanwhile, pedestrian deaths increase because 'noise' was not the problem or solution. Regardless, NHTSA has a process that starts with an investigation.

Perhaps we should start with the source documents: 2021 TESLA MODEL Y 7-SEAT SUV RWD | NHTSA

April 25, 2024 NHTSA ACTION NUMBER: RQ24009OPEN INVESTIGATION
Recall 23V838 Remedy Effectiveness
NHTSA Action Number: RQ24009

Components ELECTRICAL SYSTEM

Opened From: April 25, 2024–Present

Summary



The Office of Defects Investigation (ODI) is opening a Recall Query to assess the remedy adequacy of Recall 23V838. On December 12, 2023, Tesla filed a Defect Information Report (Recall 23V838) applicable to all Tesla models produced and equipped with any version of its Autopilot system, which Tesla described as an SAE Level 2 (L2) Advanced Driver Assistance System (ADAS). Autopilot is the simultaneous engagement of Tesla’s Traffic-Aware Cruise Control (TACC) and Autosteer. In describing the safety defect, Tesla’s Defect Information Report (DIR) explained that “the prominence and scope of the system’s controls may be insufficient to prevent driver misuse,” and Tesla committed to the deployment of a multipart remedy aimed at improving system and engagement controls and reducing mode confusion.



EA22002 (upgraded from PE21020) was opened to investigate whether Tesla’s Autopilot contained a defect that created an unreasonable risk to motor vehicle safety and involved extensive crash analysis, human factors analysis, vehicle evaluations, and assessment of vehicle control authority and driver engagement technologies. The work conducted in these investigations aligns with Tesla’s conclusion in its 23V838 recall filing. During EA22002, ODI identified at least 13 crashes involving one or more fatalities and many more involving serious injuries in which foreseeable driver misuse of the system played an apparent role.



Tesla filed Recall 23V838 to address concerns regarding the Autopilot system investigated in EA22002. Following deployment of the remedy in Recall 23V838, ODI identified concerns due to post-remedy crash events and results from preliminary NHTSA tests of remedied vehicles. Also, Tesla has stated that a portion of the remedy both requires the owner to opt in and allows a driver to readily reverse it. Tesla has also deployed non-remedy updates to address issues that appear related to ODI’s concerns under EA22002. This investigation will consider why these updates were not a part of the recall or otherwise determined to remedy a defect that poses an unreasonable safety risk.



ODI is therefore opening this Recall Query investigation to further evaluate the adequacy of the remedy for recall 23V838.



36 Affected Products
3 Associated Documents
​

It bears monitoring.

Bob Wilson

 

Found the squeaking door: Federal Regulators Investigating Tesla’s Autopilot Recall Fix - Consumer Reports

By Keith Barry
Published February 12, 2024 . Updated April 26, 2024

. . .
According to documents provided by NHTSA, the new investigation comes after the agency looked into crashes of vehicles that received the software update and did its own testing of vehicles running the new software. In addition, NHTSA says it will also look into why Tesla owners must opt into a portion of the software update, why drivers can undo it, and why later software updates to Autopilot were not included in the recall.
. . .​

Sounds like I need to submit a FAR and ask for a Congressional inquiry letter.

LATE UPDATE:

This is what Consumer Reports claims:

Driver Attention and the In-Cabin Camera

What Tesla says: “Increased the strictness of driver attentiveness requirements when using Autosteer and approaching traffic lights and stops [sic] signs off-highway.”

What CR found: If your hands are off the wheel and you’re going through a signalized intersection without another car in front, or if a traffic light turns from red to green, Autopilot will now display a message on the center screen that says, "Apply slight turning force to the steering wheel."

However, Autopilot still lacks an effective direct driver-monitoring system (DDMS), which we believe is essential to the safe operationof ADA systems. Although the in-cabin camera is capable of determining if the driver is looking away, the camera can be covered entirely and Autopilot will still work on all road types without warning the driver that the camera is blocked.

Additionally, when we covered the camera and kept one hand resting on the steering wheel, the vehicle did not limit Autopilot use or give any warnings to pay attention. According to Funkhouser, the driver could be asleep or completely distracted and the car wouldn’t warn them as long as they are holding the wheel.



Tesla's in-car camera
Photo: John Powers/Consumer Reports

New Warnings and Notifications
What Tesla says: “Improved visibility of driver monitoring warning alerts on the touchscreen by increasing the text size and moving the notifications to a more prominent position.” These changes apply to the Model 3 and Model Y only.

What CR found: Warnings such as "Apply slight turning force to steering wheel" and "Please pay attention to the road" got slightly larger and moved from the bottom to the top of the screen. Due to the design of the Model 3 and Model Y, the only place to display warnings is on the center screen. As a result, these messages actually draw the driver’s eyes to the center screen inside the car rather than toward the outside roadway.



Tesla's new Autopilot activation menu
Photo: Kelly Funkhouser/Consumer Reports

Engaging Autopilot
What Tesla says: “Added option to activate Autopilot features with a single click, instead of two, to help simplify activation and disengagement.”

What CR found: Autopilot is Tesla’s name for the driving mode in which both adaptive cruise control (ACC) and lane centering assistance (LCA) are working at the same time. Previously, drivers could engage ACC alone with a single pull of the steering wheel stalk. A second pull would also activate LCA. Drivers now have the option to activate Autopilot with a single pull of the steering wheel stalk instead of a double pull.

Our testers found that this change could present a challenge: With single-pull activation, drivers are unable to use ACC alone without LCA unless they change a setting in the vehicle menu. If you are using Autopilot and manually steer the vehicle to avoid a pothole or an obstacle, it disengages both LCA and ACC at the same time, and the vehicle stops maintaining a set speed. With double-pull activation, manually overriding the steering only disengages LCA, and not ACC.

“No other system on the market turns off ACC when you apply steering to avoid a pothole,” says Funkhouser. “It makes you unpredictable to other drivers and creates unnecessary steps to reactivate the system again—rather than just reactivating LCA, you also have to reset the ACC speed, too.”

Autopilot Strikes and Suspensions
What Tesla says: “Introduced a Suspension Policy that will restrict Autosteer usage for one week if improper usage is detected. Improper usage occurs when you, or another driver of your vehicle, receive five ‘Forced Autopilot Disengagements.’”

What CR found: FSD always had a five-strikes-and-you’re-out policy. Now, it also applies to Autopilot. The strike count is combined for the systems.

Our Model Y currently has two strikes: We got one while testing the limits of the in-cabin camera by looking away from the road while driving. We got the second when the tester was looking at the road and their hands were actively moving the steering wheel, but the system didn’t recognize the input. “The ‘strikes’ policy is not an effective method for teaching safe driving behavior because the system isn’t accurately detecting driver engagement,” says Funkhouser.​

So let me address my experience with the CR cited, specific points:

• Driver Attention and the In-Cabin Camera - I wear sunglasses during the day to protect my eyesight but also let me check the navigation map to find chargers. I have already gotten one strike at night, no sunglasses, while trying to find a charger returning from the eclipse.
• New Warnings and Notifications - there has been some changes but they look more like edits than a change of content.
• Engaging Autopilot - one or two clicks no effect. I see an option to make it a double click but mostly, a big YAWN. This is the turn signal stick that has two depression ranges: (1) light enables the turn signals, and; (2) stronger over detent, enables Full Self Driving, my default. Drivers can set to either single or double taps.
• Autopilot Strikes and Suspensions - I have one from looking too long at the navigation map for a charger. Four to go but the significantly improved FSD driving makes it easy to avoid future strikes. In the past, they were too easy.

Bob Wilson

 

I'm using Mr Google's copy. <GRINS>

The price matches the value.

Bob Wilson

 

About once a year, in January/February, I browse a copy on a news stand and after checking what they claim for used cars, compare it to my experience:

• 2019 Tesla Model 3 Standard Range Plus
• 2017 BMW i3-REx
• 2017 Prius Prime (previous)
• 2014 BMW i3-REx (previous)
• 2010 Prius (previous)
• 2005 Prius (previous, seldom reported)

I typically look at their table of metrics, the bubble chart, and their score. So far, I keep seeing the same problem of what with how they weigt the table vs overall score. When they publish how they calculate their score, I'll reconsider a subscription.

I have more respect for Motor Trend, Car and Driver, and Edmunds. Their reports document what they measure better than the Consumer Reports bubble chart. I also like the IIHS reports. Then there are my 'lying eyes.'

If I have to choose between some 3d party opinion, not backed by hard metrics, and the kind of benchmarks I run, mine will always win. Mine are reproducible and parallel the EPA metrics.

So my ranking:

• My benchmarks and Argone Labs and Munro & Associates
• EPA and IIHS
• Car and Driver, Motor Trend, and Edmunds
• Consumer Reports and YouTube Videos
• advertisements and bar talk

Bob Wilson

 

deflection
.

 

Seems to me the simplest solution would be for nhtsa and Tesla to agree on the changes before implementing them

 

think of NASA .... g'vmt agency .... constipated, blind regarding oversight ... top heavy with administrators Etc. Those kinds of Dynamics led to Tesla being successful with SpaceX .... NOT being like one of these agencies.
.

 

If car companies waited until they had an agreement from NHTSA we would still be using crank starters

 

Maybe when it comes to piloting 4000+lb death machines, we shouldn't leave the safety of those in and around them up to whatever some software developer at Tesla decides is good enough. (I'm a software developer so I feel very comfortable in saying this.) Especially since quality assurance is a corner that's regularly cut industry-wide.

 

Part of the reported issue is that Tesla let the owner opt out of a safety recall, and do so after the install. While owners could always choose not to take a car to the dealer, or even undo a recall fix, they didn't have the manufacturer helping them with that. It's like VW installing a switch to turn off the emissions fix on those diesels.

On another note, I'm pretty sure the recall process already involves discussions between the two on what the fix will be.

 

you mean in essence - the article feeds into the typical stock shorting - Tesla derangement syndrome folk? Far be it from the OP to be part of that bunch of characters.

.