Driver Assist Is More Ubiquitous But Still Not Good Enough

Over the past decade, advanced driver assistance systems (ADASs) have gone from being limited to expensive premium vehicles to standard equipment on the most affordable entry-level models. However, despite the near ubiquity of ADASs on new vehicles in 2022, the industry has yet to demonstrate the efficacy of the technology. Given the preliminary 2021 road safety data just published by the U.S. National Highway Traffic Safety Administration (NHTSA), it’s clear that drivers, industry, and policymakers need to do better. 

Over the past several years, AAA has conducted multiple tests of vehicles with a variety of ADAS features and has consistently found them wanting. The most recent test conducted at the GoMentum Station test facility in California involved a Tesla Model 3, Hyundai Santa Fe, and Subaru Forester. Each vehicle is equipped with a system that falls within the SAE Level 2 category for systems that control both speed and steering, but all three (AutoPilot, Highway Drive Assist, and EyeSight) are hands-on, eyes-on, brain-on systems that require the driver to remain in control and are not fully automated. 

In tests evaluating an approaching slower vehicle or cyclist in the same lane, all three vehicles managed to detect the target and slow down enough to avoid a collision. However, in tests where the cyclist turned in front of the test vehicle or an oncoming vehicle crossed the divider into the lane, all three vehicles failed repeatedly and impacted the targets. 

Although these AAA tests were done in sunny, daylight conditions, other ADAS tests have shown even worse results at night or other adverse conditions. My own personal experience evaluating a wide range of vehicles with many types of ADAS from numerous manufacturers is consistent with this experience. These systems tend to be wildly inconsistent and often ineffective in the conditions where drivers need augmentation the most. 

Preliminary data from NHTSA for 2021 shows that nearly 43,000 people died on American roads last year, the highest since 2005. That’s a 10.5% increase from 2020, which had already seen a significant increase from prior years despite reduced vehicle miles traveled during the pandemic. 

Reducing crashes and traffic fatalities will not be a simple problem to solve. Although the technology available across the vehicle fleet did not fundamentally change from 2019 to 2021, what did change was driver behavior, with more people taking more risks such as driving faster. The behavior problem can be addressed by better education and enforcement. But technology can help as well and should. 

Most current ADASs rely on relatively low resolution cameras and, in some cases, radar as well. Although these are adequate for applications such as lane following or adaptive cruise control, they have reached their limits as conditions become more challenging. FLIR has demonstrated the ability of its thermal imaging cameras to enhance the ability of ADASs to detect vulnerable road users in many conditions. Luminar has done public demonstrations of its lidar as part of automatic emergency braking, significantly outperforming Tesla’s AutoPilot. 

Automakers are beginning to add more capable sensors that can enable improved perception capabilities. Volvo, Polestar, BMW, Mercedes-Benz, GM, and others are adding lidar; Fisker is one of the first to add imaging radar. Higher resolution cameras are also going on new models. 

New, more stringent evaluations of ADASs are coming to tests run by EuroNCAP, the Insurance Institute for Highway Safety, and NHTSA starting in 2023. Automakers like to use these safety ratings in their marketing and they are working to improve ADAS performance. Automaker efforts are laudable; however, to reverse the disturbing trend behavioral and policy changes are needed.