Not all level 2 driver assists are equal, IIHS finds after testing

(Source: arstechnica.com)

Not all adaptive cruise control and lane keeping systems are created equal. That should be obvious, but in case it isn’t, recent tests by the Insurance Institute for Highway Safety will serve as proof. On Tuesday, IIHS published the results of some road and track testing with several different makes of vehicle, measuring how well each was able to stay within its lane around corners and up hills.

Further Reading

From Audi to Volvo, most “self-driving” cars use the same hardware“We zeroed in on situations our staff have identified as areas of concern during test drives with Level 2 systems, then used that feedback to develop road and track scenarios to compare vehicles,” said IIHS Senior Research Engineer Jessica Jermakian. By “level 2,” Jermakian means vehicles that are capable of steering, braking, and accelerating for themselves but only with an engaged human driver behind the wheel who is responsible for providing situational awareness.

The vehicles (and advanced driver assistance systems) that IIHS tested were a 2017 BMW 5 Series (Driving Assistant Plus), 2017 Mercedes-Benz E-Class (Drive Pilot), 2018 Tesla Model 3 (Autopilot 8.1), 2016 Tesla Model S (Autopilot 7.1), and 2018 Volvo S90 (Pilot Assist). IIHS notes that each had previously been scored highly for their automatic emergency braking (AEB) systems. A series of track tests was used to establish how each system coped with avoiding collisions with a stationary vehicle, then repeated road testing was conducted to get a sense of real-world behavior in traffic and how well each car maintained position within its lane.

Adaptive cruise control and braking for stationary cars

In the first series of tests, all of the vehicles successfully avoided a collision when confronted with a stationary car while traveling at 31mph, with the Volvo braking the most aggressively, similar to its AEB activation. (Interestingly, both Teslas struck the stationary vehicle in the AEB test.) Encouragingly, all of the vehicles avoided hitting a stationary object when following a lead car that suddenly changed lanes—this exact “cut-out” scenario was flagged as being outside the operational domain for most advanced driver assistance systems (ADAS) in the NHTSA investigation into a fatal 2016 Tesla crash in Florida.

The performance wasn’t quite as good in the real world, unfortunately. IIHS found that, with the exception of the Model 3, each car had occasions when it failed to stop for a stationary vehicle ahead (which is why it’s so important for the human driver to always be paying full attention to the road). The Model 3 actually suffered from the exact opposite problem; instead of false negatives it was sensitive to false positives, unexpectedly slowing on 12 occasions, seven of which were for tree shadows on the road. While that’s probably annoying for the driver, I’ll gladly take a false positive over a false negative under these conditions.

Lane keep assist

The five vehicles were also tested to see how well their lane keep assist functions worked on curves with three different radii. Again, the Model 3 stood out: it was the only car to stay within its lane every time the tests were repeated. IIHS noted that the BMW, Volvo, and Mercedes-Benz did not provide “enough steering input on their own to consistently stay in their lane, often requiring the driver to provide additional steering to successfully navigate the curve.”

Real-world driving with curves and hills tripped up all of the vehicles other than the Model 3, as you can see from the table above. BMW’s system appears to be the most flustered by elevation changes, although hills also caused the Model S to stray outside its lane relatively often. However, it’s very important to note that IIHS says it is not ready to start officially consumer rating the performance of these various systems. For that, much more testing is needed.

Further Reading

Why emergency braking systems sometimes hit parked cars and lane dividersWe will certainly welcome it when more of that data becomes available. On several occasions, readers have asked us to put together a group test of different ADAS implementations, but it’s the sort of project that quickly requires a lot of time and resources if you want to do it properly.

Perhaps the best attempt so far was the one conducted by our friend Alex Roy over at The Drive back in 2016. But two years is a long time in this space. Even within a single OEM’s range, it’s possible to find quite different performance between different models, and the older the system, the less capable it is, by and large. In the meantime, whether you drive a Tesla or a Toyota, if you’re going to use adaptive cruise and lane keeping, just remember: it’s always your job to keep your eyes on the road.

More Info: arstechnica.com

Leave a Reply