Featured IIHS automation safeguard ratings: only TSS 3.0 rated acceptable, Tesla and most others get an F

Ok. But the blame can also be attributed to poor street designs. People drive faster a lot of times because it feels like the right thing to do, which is what you feel when you're on a big wide open street. Make streets cramped and windy, and people will automatically feel like driving slower even without speed bumps. But put a 15mph sign on a continuous 50ft or 60ft wide or wider street and people will naturally speed. Take the parking elsewhere, make it just wide enough for vehicles to fit, don't make streets straight, get rid of double and triple lanes unless completely removed from neighborhood areas.

You can also add removing points of intersection into that safety equation. If every street had a median right down the middle then you wouldn't have people turning left through oncoming traffic (which also adds to not paying attention to pedestrians) in order to get to a business or driveway. Add roundabouts so people have an easier way of turning around and coming back to those businesses on the other side of the street.

Roundabouts also slow down traffic before they come together, minimizing the chances of a high-speed collision, unlike 2way stop signs and street lights that have oncoming traffic flying through thinking nobody is going to pull out in front of them. You can also run a stop sign or a red light, but it's much more difficult to run a roundabout.

And yes, a lot of this I am suggesting for neighborhood streets too. The more cramped and windy, the slower people will go. It should be a law to put a median in the middle of every street.

 

I haven’t tried dashboard dining yet and am not much of a groomer. I comb my hair once in the morning.

 

Not unless you're deliberately trying to skew the comparison

 

But the difference goes beyond having lidar. Teammate sounds like Tesla's FSP in function. You got to spend $115k on a LS 500h to get it.

Then the IIHS tested multiple systems from the same manufacturer with them getting different rankings. The Lexus Dynamic Radar Cruise Control with Lane Tracing Assist (LSS+ 3.0) got lower safeguard ratings; it's below the median on that list.

 

Yes, and European NCAP has rated the 2022 Model Y and Model S having the best Safety Assist (what ADAS is based upon) at 98%. It's the highest for the years 2023, 2022, 2021 and 2020.

Euro NCAP | Latest Safety Ratings

(On a phone, view in Landscape and sort by the last field - Safety Assist).

 

I just finished "Safeguards for Partial Driving Automation Test Protocol and Rating Guidelines" Version I, March 2024. Overall, a good first draft but I've annotated my hardcopy with these concerns:

• "Core rating" pp3 - no mention of time limits used within the document that range from 5 to 30 seconds. We really need to see what time ranges mean in scoring.
• "geofenced fenced" pp4 - weakly covered and no mention of defective maps.
• "normal or default" pp5 - Tesla has three degrees of Automated Emergency Braking, Late, Medium, and Early. This is used to tune out or increase false braking. An omission.
• "grade 90 cheesecloth" pp7 - now I know how to make my 'Spiderman' goggles to use at night. They had correctly identified sunglasses as a defeat device in daylight.
• "5, 10 sec" pp13-14 - repeat of pp3 problem as no time scale justification or rational.
• "keeps both hands" pp15 - NO, single hand driving is OK. No mention of hands on wheel detection mechanism.
• "picks up the [cell phone sized, rjw] block" pp17 - no discussion of hands free cell phone operation or relocation but not use.
• "Automated lane change" pp19 - writing style changed but legit as Tesla "minimize lane change" is not working.
• "Test 8a, 8b" pp19-20 - roughly describes behavior in traffic jam or work zone moderated operation. Poor retationship to common traffic problems.
• "Cooperative steering" pp 21-22 - requires manual only operation with no accounting for Tesla having automated lane shifting. BIAS
• "repeated attempts" pp23 - Nag facist, too strict.
• "deactivating AEB" pp25 - Tesla has three mode, late, average, early, which are used to minimize false braking. Ignorance of Tesla AutoPilot and Full Self Driving.
• "monitor driver body, head, and hands" pp27 - no mention of variable steering torque.
• "a system lockout" pp28 - no mention of duration measured in time, trip, or weeks. Nag facist.
• "Automated lane change . . . ACC auto-resume" pp29 - automated lane changes to stay on route, minimize passing lane hogging, and passing slower traffic is necessary.

I give IIHS an "A" for having a reasonably well written standard testing document. There are a few areas we disagree or they are wrong (i.e., traffic jams and work zones.) More clarity on the time limits is needed. As a version 1 standard, a good step forward over the usual crap reviews. My biggest take away, I really like the cheesecloth defeat device or 'Spiderman goggles."

Bob Wilson

 

Late thought, editing a document has something to mark up. But omitted in "Safeguards for Partial Driving Automation" are these:

• Phantom braking - often caused by defective navigation maps that override the posted speed limit. These can cause sudden speed decreases. But to test it, you need a defective map which is not a trivial problem. Call it the silent killer. They can also be induced by the Automatic Emergency Brake system reacting to cross traffic stopped at an intersecting road.
• Hands free phone - mechanical operation of a cell phone is mitigated by "hands free" operation. This should be evaluated and scored in the test.
• Merge and exit lanes - typically caused by absence of lane divider lines or dashed lines, this can cause the car to "split the difference" in worst cases, appear to drive toward the far corner ditch.
• STOP, YIELD, TRAFFIC LIGHTS - an obvious omission.
• Geofenced vs not - the ratio of geofenced miles versus not fenced should be scored. The higher the ratio, the better the score.

I have five years of AutoPilot and approaching two years of Full Self Driving experience. So I have some insights to what was not in Version 1.

FYI, the list of omitted tests might change the scores. Regardless, I've already voted with my wallet.

Bob Wilson

 

LSS+ 3.0 came out in 2021 for the 2022 NX. The IIHS tested the systems on the 2022 to 2024 LS, which the Lexus site lists as having LSS+ 3.0. Lexus Safety Technology | Lexus.com

Perhaps TSS+ and LSS+ three have different levels of safeguards. You can't take that to assume TSS+ 3.0 has the same ones as the Lexus Teammate, which has greater ADAS capabilities than TSS+ 3.0.

 

Interesting, NCAP gave the Lexus NS a Safety Assist rating of just 91%, lower than what the Model Y and Model S got for the same year (98%).

Official Lexus NX 2022 safety rating

Official Tesla Model S 2022 safety rating

Official Tesla Model Y 2022 safety rating

 

It should also be pointed that they tested Tesla's Autopilots 2023.7.10, which is way before Tesla Over the Air updated the Autopilot and FSD stacks to NHTSA requirements in 2023.44.x. It monitors drivers attentiveness way more now and will give you "strikes" if it has to disengage because of driver's negligence. After five suck strikes, AP disables for a week for anyone driving the vehicle. I'm currently sitting at zero strike because I don't abuse the system.

 

I've been reading a bit more about the Lexus Teammate, the only system that scored anything other than poorly.

It seems to have a pretty long list of features, more than I've heard anyone mention in relation to the features available in any Prius.

Can you clarify what you mean by "tested as Lexus teammate?"

From what I've been reading, it's simply not the same as Toyota's TSS 3.0. A lot is shared, there is obvious overlap... but they aren't identical.

 

Prii with Traffic Jam Assist may come close to Teammate. That installs a set of driver monitoring cameras that aren't part of the base TSS+ 3.0.

The IIHS wants the ADAS to disengage for the rest of the trip after after one strike. If a system doesn't do that, it fails that part of the test. I understand they need a testable metric. It is just some of these requirements tests don't really apply. We should all be wearing seatbelts, but how is that a safeguard to proper use of ADAS? I would like to know the reasoning for them, and any evidence for selecting the criteria they did.

I do think having this rating a good idea. It is as important as how the ADAS works. Just a shame many reports are acting like it reports the latter.

 

maybe you don't
maybe you do .....




.

 

I can't vet for FSD but Autopilot will disengage for the remainder of the trip. It's been like that since at least 2021. You have to stop and park before you can re-engage it, that's assuming you haven't reached that five strikes (new since 2023.44.x) because then, you won't be able to re-engage Autopilot for a week.

 

Seems even rarer of a situation.

The current Lexus LS is available with different safety packages in different trims, and I've found a claim that they only sold 107 cars with the full Teammate system in 2023. Haven't found a good confirmation of that though.

Just an interesting twist on the test results. Only one system didn't fail badly, and that particular system is only rolling around in a very few cars.

Would be statistically embarrassing for any of them to have an upsy-daisy.

 

Lol, plain sunglasses is enough. I think Ford or GM have IR camera that sees through sunglasses and at night but Teslas don't.

 

Well, Teammate is only available with the LS hybrid, which is $115k, or around $30k more than the non-hybrid. Not surprising there isn't many around.