Tesla Model 3 Outperforms In New IIHS Level 2 Autonomy Tests


However, it’s not perfect.

With drive assistance features becoming standard in many vehicles, it’s good to remember they’re not all the same. Some work better than others, and the Insurance Institute for Highway Safety is devising a series of tests to evaluate these systems in various driving situations. The focus of the tests is Level 2 “autonomy” as defined by SAE International, which includes adaptive cruise control (ACC) and active lane-keeping (ALK).

In IIHS’s research, the company tested five vehicles: 2017 BMW 5 Series with “Driving Assistant Plus,” 2017 Mercedes-Benz E-Class with “Drive Pilot,” 2018 Tesla Model 3 and 2016 Model S with “Autopilot” (software versions 8.1 and 7.1, respectively), and 2018 Volvo S90 with “Pilot Assist.” The Tesla Model 3 performed best through the bevy of tests – however, it wasn’t perfect.

The cars were put through four different scenarios to test their ACC. The first involved driving at 31 miles per hour toward a stationary vehicle with ACC off and autobrake turned on. Only the two Teslas failed, hitting the stationary target. The same test was then performed with ACC on, and the Tesla Model 3 slowed with gradual decelerations. All vehicles passed this portion.

A third scenario had the cars follow a lead vehicle that slowed to a stop and then accelerated. Every car performed well in this test. The final test had the test cars follow a lead vehicle that changed lanes to reveal a stationary vehicle in the test vehicle’s path. The vehicles had about 4.3 seconds before colliding with the stationary vehicle. However, all the test cars performed well with none of the vehicles striking the stationary vehicle.

Where the Tesla Model 3 truly outshined its competition was in the hill and curve tests for ALK. Here, the IIHS conducts six tests on three different sections of curved roads. Only the Model 3 stayed within its lane through all 18 trials. To test how the ALK of all five vehicles performed on hills, the IIHS mapped out a course on three hills with different slopes, running six different tests on each hill in each vehicle. Here, the Model 3 had just one deficiency, touching the centerline once in 18 tests.

The Model 3’s competitors had various levels of success with the ALK tests. The BMW, Mercedes, Model S, and Volvo all went over the centerline during both the hill and curve test, with the Model S crossing the most at 12 times when being tested on the hills. The Volvo crossed the line the most at eight times during the curve test. Some of the vehicles, such as the 5 Series, E-Class and S90 had the ALK system disengage during the tests, with the 5 Series system disengaging the most.

As these systems infiltrate new cars, drivers need to understand their limitations and the differences between makes and models. It will be up to agencies such as the IIHS and the National Highway Traffic Safety Administration to design tests that both inform consumers and keep them safe because driver assistance features aren’t going anywhere. However, it’ll be years before these systems are foolproof.

“We’re not ready to say yet which company has the safest implementation of Level 2 driver assistance, but it’s important to note that none of these vehicles is capable of driving safely on its own,” David Zuby, IIHS chief research officer, says in the study. “A production autonomous vehicle that can go anywhere, anytime isn’t available at your local car dealer and won’t be for quite some time. We aren’t there yet.”

Categories: Tesla

Tags: , ,

Leave a Reply

32 Comments on "Tesla Model 3 Outperforms In New IIHS Level 2 Autonomy Tests"

newest oldest most voted

GM Supercruise would own this space if they’d roll it to more cars.

Dream on troll.

Don’t have to dream, I’ve read the reviews. The SuperCruise interface is better (watches the driver’s eyes vs requiring hand on wheel) and it uses pre-mapped roads to avoid the crashes that happen with Autopilot.

Name calling is always the last resort of those who don’t have facts on their side.

From what you stated, GM Supercruise is the worst, it can’t even drive on the test track. The owners can only drive on specific roads. It’s like living in the bubble.

A bubble that includes all major highways.

At some point systems like Autopilot will be able to independently “see” the road and deal with unmapped roads. That’s probably at least 5 years away. Until then, pre-mapping interstate highways is the safest option for semi-autonomous driving.

In that case Tesla wins again. It was first to include autonomous and has been mapping the roads of every one of their for more moles than GM by hundreds of millions of miles.

The mapping can’t be done by regular cars. High def maps of roads using special equipment are required. GM has those maps.

All major highways are the easiest to navigate. Backroads, school campuses, parking lots, private roads etc is where it gets challenging. Saying Cadillac is best where it’s easiest isn’t saying much.

None of the systems, including autopilot, are supposed to be used on secondary roads. Interstates only.

It wasn’t in this test. Have it run this test, then you can have Supercruise stare into my eyes all day. You can’t make a statement without proof or even evidence.

I’ve read the professional reviews. All the pro reviewers prefer way supercruise tracks if the driver is paying attention.

The tests are not on pre-mapped roads.

What does ” it uses pre-mapped roads ” mean? For example, does that include the IIS test track? How about the recently completed, reworked Parkway between Golf Road and Mountain Gap?

How are roads “pre-mapped”? Who pays for it? How are they kept current when a storm washes out a road or bridge?

Google supercruise and you can read the reviews. GM paid for highly detailed maps covering 400,000 miles of roads in the US/Canada. The mapping detail can only be achieved with special equipment, not autopilot cameras, so Tesla won’t be able to learn it using their deployed fleet.

I wonder if the difference between Tesla vehicles was primarily the car or primarily the autopilot version. In other words, would a Model S with the latest version of autopilot do as well as the Model 3?

I wonder why they did not test a Model S with the newer 8.1 version of Autopilot that matches what the Model 3 has so that results would be the same???? Also knowing that Autopilot V9 is to be released this month, why IIHS did not wait a couple weeks to test since this version is supposed to have as Musk explains “With V9 we will begin to enable full self-driving features.” Exactly what the IIHS says none of the cars will be able to do for ‘quite some time.’

It does not have the newer hardware that the M3 has.

Correct headline should be: Tesla Autobrake failed miserably ,while all other passed the Test

I thought EVANNEX penned this article based off the headline.

How did the Bolt break out in this test?

That is the absolutely most important part.

Where does it say in the test that they “failed miserably”? Teslas passed all the autobrake scenarios except just the one at 31 mph without ACC on. And it says they hit the stationary target but don’t say how fast they were going when they hit. If they were slowing down and hit the target going 10 mph or 5 mph then it did its job which is to save lives. You will still have a wrecked car however. Both Teslas passed the 31 mph test with ACC turned on.

From the IIHS report: “Out on the road, engineers noted instances in which each vehicle except the Model 3 failed to respond to stopped vehicles ahead. Jermakian recounts her experience with the E-Class on U.S. 33 near the IIHS-HLDI Vehicle Research Center (VRC). Traveling about 55 mph with ACC and active lane-keeping engaged but not following a lead vehicle, the E-Class system briefly detected a pickup truck stopped at a traffic light ahead but promptly lost sight of it and continued at speed until she hit the brakes.”

The difference is Tesla will fix that tomorrow with an OTA update.

does anyone know where EAP ends and FSD starts for tesla? level 2 ? level 3? i am really happy with my current EAP on model 3. it can be lot smoother and less erratic when cars merge into the lane or come close to.

It’s hard to tell. IIRC the upcoming on-ramp to off-ramp feature, which is the first to venture into level 3, is still supposed be part of EAP; on the other hand, they said before that FSD would use more cameras than EAP — but I don’t think it would make sense to limit the number of cameras used for on-ramp to off-ramp…

(⌐■_■) Trollnonymous

No thanks on AP. That’s one feature I will not pay for.

Have you driven with AP?

AP #9 is going to be even better, though it’s difficult to see how they could improve on near perfection.
I suppose it’s more aimed at the complicated on/off ramps stuff, which is another level of difficulty.
Good work on the auto pilot, further extending Tesla’s lead in that venue, over all competitors by a substantial margin.

Did anyone else catch the somewhat of a hit-piece by NBC Nightly News today regarding autonomous vehicles? It painted the technology in a very bad light. They pretty much talked about Mercedes, BMW, and Tesla; and they really didn’t give any good context or facts, just a bunch of fear-mongering IMO. They could have talked about all the autonomous Bolt EVs driving around San Francisco. Or the autonomous buses at the University of Michigan. But no. They just showed pictures of crashed Tesla’s without any back-story, with random people saying how they would never trust autonomous cars.

That is what they do best (for their ratings )

It’s unfortunate that the new Leaf with ProPilot was not included in this testing. I’m confident it would have scored just as well as the Model 3. I’ve personally used it on winding 2-lane country roads and it works really good.

I guess the IIHS is only thinking that “luxury” cars have Level 2 autonomy features like lane assist. However there are many other lower cost cars that do also including the Leaf.

I hope they revisit this testing to include a wider sample group.

Calm my friends as this technology is new and rapidly changing. Ten days ago, Tesla revealed a custom board that handles 2,000 frames per second versus 200 in the ones tested at IIHS. A board that can be retro-fitted to earlier Tesla cars. That along with version 9 of the software.

Testers can only use what is currently available and almost as soon as they finish one set, they’ll use lessons learned to make the test harder while newer hardware and software is released.