Adaptive cruise control, active lane-keeping, and other automated features can all be useful in preventing crashes. But the Insurance Institute for Highway Safety is reminding drivers that they should continue to be vigilant behind the wheel, since these systems aren't yet reliable enough to take over driving tasks.

IIHS recently completed several tests on both public roads and tracks in its efforts to develop a ratings system for advanced driver assistance systems. The systems showed variable reliability, exhibiting behaviors ranging from overly cautious braking to veering toward the shoulder when lane markings could not be detected.

Advanced driver assistance programs are part of automakers' efforts to develop fully self-driving vehicles, which are defined as Level 5 on a scale developed by SAE International. IIHS looked specifically at vehicles with a combination of adaptive cruise control and active lane-keeping systems, which are considered to be at Level 2. While the systems can assist with driving tasks, human drivers are still in charge of safely operating the vehicle.

The testing comes in the midst of greater scrutiny of advanced driver assistance programs, particularly the Autopilot system developed by Tesla. Earlier this year, two American consumer rights groups accused the system's name of being "deceptive and misleading," causing several drivers to rely too heavily on the system.

Tesla has stressed that drivers need to remain focused while using the system, a small number of crashes and incidents have occurred while Autopilot was activated. The National Transportation Safety Board said a fatal 2016 crash involving a Tesla was due in part to the driver's "inattention due to overreliance on vehicle automation," and a British man had his license suspended earlier this year after he was caught riding in the passenger seat while the Autopilot system was activated.

"Designers are struggling with trade-offs inherent in automated assistance," said David Zuby, chief research officer at IIHS. "If they limit functionality to keep drivers engaged, they risk a backlash that the systems are too rudimentary. If the systems seem too capable, then drivers may not give them the attention required to use them safely."

"The new tests are an outgrowth of our research on Level 2 autonomy," said Jessica Jermakian, senior research engineer at IIHS. "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."

Researchers used a 2016 Tesla Model S and 2018 Tesla Model 3, each equipped with Autopilot, as well as a 2017 BMW 5-series with Driving Assistant Plus, a 2017 Mercedes-Benz E-Class with Drive Pilot, and a 2018 Volvo S90 with Pilot Assist for the testing.

Adaptive cruise control

Adaptive cruise control systems adjust a vehicle's speed to maintain a following distance from a lead vehicle. The IIHS tests evaluated how the different systems reacted in a variety of common scenarios.

Each vehicle decelerated smoothly when the lead vehicle slowed to a stop before accelerating. Most vehicles also had a smooth deceleration when the adaptive cruise control system was set to differing following distances, although the S90 braked more forcefully. This vehicle also decelerated more aggressively in a test where a lead vehicle changed lanes to reveal a stationary inflatable vehicle target ahead, with 4.3 seconds to react before a collision.

Researchers noted how track tests can differ from real world performance, citing the reaction of adaptive cruise control to stopped vehicles as an example. Each of the test vehicles' owners manuals warn that the system may not brake for vehicles that are already stopped when they come into sensor range. Jermakian said she personally experienced this shortcoming when driving an E-Class near IIHS headquarters, when the Drive Pilot system did not slow down for a pickup truck stopped at a traffic light, thus requiring Jermakian to apply the brakes herself.

"At IIHS we are coached to intervene without warning, but other drivers might not be as vigilant," she said. "ACC systems require drivers to pay attention to what the vehicle is doing at all times and be ready to brake manually."

Researchers also conducted a test of the vehicle's autobrake system when approaching a stationary vehicle target at 31 miles per hour. The Tesla models were the only ones to hit the target in the test.

In addition, the Model 3's adaptive cruise control system slowed unexpectedly 12 times in 180 miles of driving. In seven instances, the system misinterpreted shadows cast by trees as an obstacle. In other cases, the vehicle braked due to oncoming traffic or vehicles crossing the road a safe distance ahead.

"The braking events we observed didn't create unsafe conditions because the decelerations were mild and short enough that the vehicle didn't slow too much," said Jermakian. "However, unnecessary braking could pose crash risks in heavy traffic, especially if it's more forceful. Plus, drivers who feel that their car brakes erratically may choose not to use adaptive cruise control and would miss out on any safety benefit from the system."

IIHS said that while current adaptive cruise control systems aren't capable of regulating speed in all traffic situations, they have still been shown to help promote safer following distances and reduce the risk of a crash. The system is often paired with forward collision warning and automatic braking systems.

Active lane-keeping

In testing the active lane-keeping systems, researchers focused on how the feature performed on hills and curves. The tests were done on an open road with no other traffic present.

Each test vehicle's active lane-keeping system uses steering assistance to center the vehicle between clearly marked lines. Sensors may also use a lead vehicle as a guide at lower speeds and when a lead vehicle is blocking the view of lane markings.

The Model 3 was the only vehicle to stay within its lane on all 18 curve tests. Similarly, the Model S stayed within the lane on 17 curve tests, although the system overcorrected on one test and crossed the line on the inside of the curve.

The other systems usually could not provide enough steering input on their own and required driver action to successfully negotiate the curve. The active lane-keeping systems on the E-Class and S90 only negotiated a curve in nine of the trials, while the 5-series completed a mere three curve tests successfully.

The 5-series' Driving Assistant Plus system disengaged on nine of the curve tests, crossed the line in three, and touched the line in one. The S90 went over the line in eight tests, while the E-Class touched the line in five tests, crossed it in two, and disengaged in one.

IIHS says active lane-keeping systems can be confused by hills, since the system might lose track of lane markings while cresting an incline. In testing the system, researchers drove the vehicles on three hills with differing inclines, completing six tests on each hill.

The Model 3 was the most successful on the hills, staying within the lanes in 17 tests and touching the line on one. By contrast, the Model S only stayed within the lane during five tests and crossed the line in 12.

The E-Class stayed within its lane on 15 hill tests; it went over the line, touched the line, and disengaged one time each on the remaining tests. The S90 stayed in the lane nine times, disengaged four times, went over the line twice, and touched the line once.

The 5-series did not stay within the lane on any of its 14 valid hill trials. The system disengaged seven times, crossed the line six times, and touched the line once.

Researchers said driver input was often required to override the system on hills, and that the systems sometimes performed erratically. For example, the Model S frequently veered into adjacent lanes or onto the shoulder while trying to find the center of the lane as it crested hills.

Systems automatically disengaged when they were unable to detect markings or when drivers intervened. Some systems automatically reengaged when they found lane markings again, while others required the driver to reactive the feature.

IIHS said the active lane-keeping feature sometimes followed a lead vehicle onto an exit instead of continuing on if it was relying on the lead vehicle instead of lane markings to keep the vehicle centered. This situation was most likely to occur when the vehicle was traveling too slowly to track the lane lines.

"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," said Zuby. "A production autonomous vehicle that can go anywhere, anytime isn't available at your local car dealer and won't be for some time. We aren't there yet."

IIHS said that while active lane-keeping systems offer more limited safety benefits than adaptive cruise control, they still have the potential to reduce single-vehicle, sideswipe, and head-on crashes caused when a vehicle departs the roadway or enters another lane.