Suddenly, Driverless Cars Hit Bumps

Recent tests by The Insurance Institute for Highway Safety on two key ADAS capabilities cast doubt on the efficacy of these technologies and thus on how soon full autonomy is likely to affect auto insurance premium. Anyone insuring automobiles is paying a lot of attention to the development of ADAS (advanced driver assistance systems) and of fully autonomous vehicles. Many of the underlying technolgies used in ADAS (e.g. cameras, radar, lidar, AI) will also be used in fully autonomous vehicles. However, the demands that a fully autonomous vehicle places on these technologies are quite different than the demands of an ADAS-equipped vehicle. ADAS-equipped vehicles will pass control to and from human drivers (or send warnings to human drivers) in various circumstances. Fully autonomous vehicles will have no hand-offs and no warnings because there are no human drivers to receive them. The Insurance Institute for Highway Safety (IIHS) recently ran a series of tests of two key ADAS capabilities: adaptive cruise control (ACC) and active lane keeping. ACC maintains a set speed and a specified distance from a car in front of the car with ACC. Active lane keeping automatically maintains the car within its current lane. See also: Autonomous Vehicles: Truly Imminent?   Vehicles with ACC and active lane keeping are at Level 2 on the SAE International scale. This is a widely recognized framework demarcating degrees of autonomy — ranging from Level 0 (no automation) to Level 5 (fully autonomous). [caption id="attachment_32988" align="alignnone" width="570"] Source: NHTSA https://www.nhtsa.gov/technology-innovation/automated-vehicles-safety[/caption] The IIHS tested five well-regarded vehicles:

• A 2017 BMW 5-series with “Driving Assistant Plus”
• A 2017 Mercedes-Benz E Class with “Drive Pilot”
• A 2018 Tesla Model 3 and a 2016 Model S with “Autopilot” (software versions 8.1 and 7.1, respectively)
• A 2018 Volvo S90 with “Pilot Assist.”

The results of these tests were reported in IIHS and HLDI publication, Status Report (Aug. 7, 2018). The results were not pretty.

• In one test on a public roadway, the Mercedes was aware of a stationary vehicle in front of it but continued without reducing speed, until the human driver applied the brakes.
• In a 180-mile test drive, the Tesla Model 3 slowed without an appropriate cause 12 times (including seven instances of tree shadows on the road).
• In testing active lane keeping on curves; the BMW, the Mercedes and the Volvo were unable to stay in their lane without the driver providing steering assistance.
• The vehicles’ active lane keeping capability was also tested when they reached the top of hills. At the top, some cars’ technologies essentially lost sight of the lane markings on the road. The BMW failed to stay in its proper lane (without driver intervention) in all 14 tests. The Volvo stayed in the lane in nine of 16 tests. The Tesla Model S swerved right and left as it attempted to locate the appropriate lane. Sometimes it also entered an adjacent lane or drove onto the shoulder.

There is evidence that ADAS technologies do reduce accidents and insured losses—here and here. See also: Autonomous Vehicles: ‘The Trolley Problem’   However, the real world test results of Level 2 technology in these five highly regarded models were certainly disappointing. Level 2 autonomy requires the driver to remain engaged and constantly monitor the environment. The key words are “remain engaged.” People, while driving, often do many things other than remaining engaged. Conclusion The shared responsibility between less-than-perfect humans and less-than-perfect technologies of Level 2 implies that either the technologies have to become intrinsically better — or they must find ways to compensate for imperfect humans. As mentioned, you cannot make a straight-line projection of elapsed time from the current state of Level 2 ADAS technology to the arrival of ready-for-prime-time Level 5 fully autonomous technology.