ARLINGTON, Va.—Driver assistance systems that can accelerate, brake and steer automatically do not perform well enough to substitute for human drivers, according to the Insurance Institute for Highway Safety.
In road and track tests of Level 2 driver assistance systems on vehicles from Tesla, Mercedes-Benz, BMW and Volvo, IIHS reported that each system had potentially dangerous weaknesses.
"We don't think any of these five systems can be relied on," said David Zuby, IIHS chief research officer. "Drivers must remain attentive when these systems are in use."
Four track-based tests revealed weaknesses in Volvo's adaptive cruise control and Tesla's automated emergency braking systems, relative to competitors from Mercedes and BMW. On-road testing highlighted shortcomings in all the systems, with IIHS test engineers reporting that every vehicle tested except Tesla's Model 3 failed to respond to stopped vehicles ahead at some point. The Model 3 displayed overly cautious braking 12 times in 180 miles of testing, with engineers reporting that seven of the incidents coincided with tree shadows on the road.
Research suggests that a trade-off exists in automated driving systems, between false negatives in which the vehicle fails to brake and false positives in which the vehicle brakes when no obstacle exists.
"It's not clear that the unnecessary braking is the direct trade-off associated with better detection of stopped lead vehicles, but it could be," Zuby said.
"We would need to do additional testing to understand that further."
On-road testing of active lane-keeping systems, which automatically keep vehicles within lane markings, also had inconsistent performance in challenging conditions. Tesla's Autosteer system fared the best in a series of six trials on three sections of curved road; only the Model 3 stayed inside the lane markings in all 18 attempts, and the Model S crossed the lane line only once. The Mercedes and Volvo active lane-keeping systems stayed inside the lane lines in nine of 17 attempts, while BMW's was successful in only 3 of 16 attempts.
In an on-road test of the lane-keeping systems' ability to navigate hills, which often cause them to lose track of lane markings, the Model 3 stayed in its lane in all but one attempt, while the Mercedes succeeded in 15 of 18 attempts. The other systems performed worse, with the Volvo succeeding in 9 of 16 trials, the Model S in 5 of 18, and the BMW failing to stay between the lane markings in all 14 attempts.
IIHS is developing a comprehensive test scheme for these systems, which is expected to be ready in about 12 months, Zuby said.
Even though the Model S and Model 3 fared well in some tests, they had the most dramatic failures during track-based testing. Both Teslas hit stationary objects in testing of their automated emergency braking performance.
Tesla's automated emergency braking system struck stationary objects in tests this year by Luxembourg's testing and standardization authority ILNAS and in NHTSA testing in 2017. Though automated emergency braking systems are not considered Level 2 because they do not automate normal driving functions, IIHS research indicates that it can reduce rear-end crashes by as much as 40 percent.
An analysis of insurance data accompanying the IIHS report suggests that Tesla's activation of its Level 2 Autopilot system led to claim reductions in line with those in other automakers' deployment of Level 1 systems, such as automated emergency braking, forward collision warning and blind spot monitoring.