Advanced driver assistance

Driverless cars aren’t here yet. More vehicles are incorporating a degree of automation with technologies such as adaptive cruise control and lane centering, but the driver will continue to share driving responsibilities for the foreseeable future. In theory, fully automated driving could eliminate the vast majority of crashes, but that level of automation won’t be parked in your driveway anytime soon.

Advanced crash avoidance features are becoming widespread. Many of today’s vehicles have technologies that monitor driver input and the environment around the vehicle and warn the driver when they detect the possibility of a crash. They may automatically brake or steer the vehicle if the driver does not act to avoid the collision.

Some driver assistance technologies are reducing crashes. Front crash prevention, lane departure prevention, blind spot detection and rear crash prevention also show real-world crash reductions.

Real-world benefits of crash avoidance technologies: summary of IIHS-HLDI findings

Front crash prevention systems are designed to intervene when the vehicle is about to rear-end another vehicle. The technology uses various types of sensors, such as cameras, radar or lidar — short for light detection and ranging — to detect when the vehicle is getting too close to one in front of it. The systems generally issue a warning and precharge the brakes to maximize their effect. Most also apply the brakes if the driver doesn’t respond.

Front crash prevention is becoming more universal and its capabilities more consistent across brands, thanks to a voluntary commitment by 20 automakers, representing 99% of U.S. light vehicle sales, to make the technology standard by September 2022. The commitment, brokered by IIHS and the National Highway Traffic Safety Administration (NHTSA), calls for vehicles to have systems with both a forward collision warning component that meets NHTSA criteria and automatic braking that achieves certain minimum speed reductions in IIHS track tests.

Many front crash prevention systems can detect pedestrians, and some also recognize cyclists and animals. These systems use advanced algorithms coupled with sensors and cameras to spot nonmotorists who are in or about to enter the vehicle’s path.

Vehicles equipped with front crash prevention are much less likely to rear-end other vehicles than the same models without the technology (Fildes et al., 2015; Isaksson-Hellman & Lindman, 2016;  Cicchino, 2017). An Institute study found that systems with forward collision warning and automatic braking cut rear-end crashes in half, while forward collision warning alone reduces them by 27%. The autobrake systems also greatly reduce rear-end crashes involving injury.

A separate IIHS study showed that automatic braking systems that recognize pedestrians cut pedestrian crashes by 27% (Cicchino, 2022).

Passenger vehicles are not the only vehicles that benefit from front crash prevention systems. Similar rear-end crash reduction effects have been found for large trucks equipped with front crash prevention systems (Teoh, 2021).

HLDI has conducted studies comparing insurance claim rates for passenger vehicles equipped with front crash prevention with claim rates for the same models without the technology. Vehicles equipped with these systems consistently show lower rates of claims for damage to other vehicles and for injuries to people in other vehicles (HLDI, 2023).

Similarly, HLDI found that Subaru’s EyeSight system with pedestrian detection cut the rate of likely pedestrian-related insurance claims by 35%, compared with the same vehicles without the system (Wakeman et al., 2019).

Even if a front crash prevention system doesn’t avoid a crash altogether, it may still reduce the impact speed, thereby making a crash less severe.

To show why reducing speed is important, IIHS conducted two demonstration crash tests at different speeds in 2013. In each test, a 2013 Mercedes-Benz C-Class ran into the back of a stationary 2012 Chevrolet Malibu. The tests illustrated what happens in a 25 mph crash when the striking vehicle doesn’t have autobrake, compared with what happens when the speed is reduced by 13 mph, the amount by which the C-Class's autobrake system reduced the impact speed in IIHS track testing. Damage in the higher speed crash test was about $28,000. The Malibu was a complete loss. Lowering the speed to 12 mph trimmed the damage to $5,700 (IIHS, 2013).

A similar speed reduction in a higher-speed crash could significantly reduce injury risk as well as vehicle damage (Kraft et al., 2009).

Front crash prevention systems with automatic braking have resulted in bigger reductions in rear-end crashes with injuries than in rear-end crashes of all severities, which suggests that these systems are preventing injuries in some rear-end crashes that aren’t avoided (Cicchino, 2017).

IIHS has rated front crash prevention systems since 2013 and began rating pedestrian detection systems in 2019.

NHTSA issued a proposed regulation in 2023 that would require all new passenger vehicles to be equipped with automatic braking that can avoid rear-end crashes with other vehicles and crashes with pedestrians (Office of the Federal Register, 2023). Pedestrian detection would be required to work in both daylight and dark conditions.

These systems use cameras to track the vehicle’s position within the lane, alerting the driver if the vehicle is in danger of inadvertently straying across lane markings when the turn signal is not activated. Some systems use haptic warnings, such as steering wheel or seat vibration, while others use audible and/or visual warnings. Some systems cause the vehicle to actively resist moving out of the lane or help direct the vehicle back into the lane through light braking or minor steering adjustments.

Lane departure warning has not brought down insurance claim rates but has reduced rates of single-vehicle, sideswipe, and head-on crashes reported to the police (HLDI, 2023; Cicchino, 2018; Sternlund et al., 2017).

This feature uses sensors to monitor the side of the vehicle for vehicles approaching blind spots. In many systems, a visual alert appears on or near the side mirrors if a vehicle is detected. An audible alert may activate if the driver signals a turn and there is a vehicle in the blind spot. Some systems also may activate the brake or steering controls to keep the vehicle in its lane.

Blind spot detection has been shown to reduce lane-change crashes by 14% (Cicchino, 2018). HLDI research has also found that blind spot detection lowers rates of insurance claims covering injuries and damage to other vehicles (HLDI, 2023).

There are many different technologies designed to help drivers back up safely. Rearview cameras display what is behind the vehicle, projecting a much larger field than is visible in mirrors or even by looking directly out the back windshield. Since May 2018, rearview cameras have been essentially required on new vehicles in order to reduce backover crashes, in which young children are frequently the victims (Office of the Federal Register, 2014).

Some camera systems, as well as systems that use radar or ultrasonic sensors, warn the driver if there are objects in the way when the vehicle is in reverse. Systems with rear automatic braking apply the brakes to keep the vehicle from backing into or over an object. A rear cross-traffic alert system detects vehicles approaching from either side that may cross the path of a backing vehicle, warns the driver, and may automatically brake to prevent a collision.

Rear automatic braking is associated with the largest reductions in insurance claims and backing crashes reported to the police of any type of rear crash prevention system (Cicchino, 2019; HLDI, 2023).

IIHS has issued ratings for some rear crash prevention systems.

Crash avoidance technologies can’t be effective unless they are used. Appropriate driver responses and acceptance of these technologies are critical to their success. If drivers don’t trust the systems or find them annoying or not useful, they may disable them. Similarly, if drivers experience warnings but don’t understand them, are overwhelmed by them, or don’t take an appropriate corrective action, then the systems will be ineffective.

Observations at dealerships of seven automakers in 2016 found that front crash prevention systems were activated in 93% of the vehicles observed that arrived for service, and nearly 100% of the blind spot detection and rear-cross traffic alert systems were turned on (Reagan at al., 2018). Activation of systems that alert (lane departure warning) and intervene (lane departure prevention) to help with lane keeping were much lower at 52%. Lane departure systems that warned by vibration were more likely to be activated than those that beeped, and lane departure prevention systems were more likely to be turned on than warning-only systems.

Drivers need to be ready and able to respond to warnings or interventions in order for them to work. Many drivers involved in lane departure crashes are asleep or otherwise incapacitated, which can limit their ability to respond to lane departure warning and lane-keeping support systems (Cicchino & Zuby, 2017;  Wiacek et al., 2017). Systems that only warn the driver are not as effective as those that act on behalf of the driver, such as automatic braking (Cicchino, 2017). However, even systems that intervene may require a follow-up response.

In addition to driver challenges, the technology itself can have limitations. For example, lane departure warning systems use sensors to register lane markings or the road edge, which may be problematic on roads that aren’t well-marked or are covered with snow. Sensors may not function well in low light or inclement weather. Some systems only work at certain speeds. While pedestrian crash avoidance systems are effective during the day or on lit roads, they struggle to detect pedestrians in the dark (Cicchino, 2022).

In driving, automation involves using radar, cameras, and other sensors to perform parts or all of the driving task on a sustained basis. One example is adaptive cruise control, which continually adjusts the vehicle's speed to maintain a set minimum following distance. Features such as automatic braking, which acts as a back-up if the human driver fails to brake, or blind spot detection, which provides additional information to the driver, aren’t considered automation under this definition.

Driving automation is not limited to vehicles that drive themselves without human interaction but includes technologies that vary in technical capability.

The National Highway Traffic Safety Administration uses definitions of different levels of automation developed by SAE International (SAE International, 2021). The levels of driving automation range from none, or Level 0, to full driving automation, or Level 5. The levels are differentiated by the roles and responsibility that the human and automation have; for example, whether a human is required to monitor the driving environment and whether, if things go wrong, the human is expected to intervene or the automated system can bring the vehicle safely to a stop.

Levels of driving automation

In vehicles in which only some elements of driving, such as steering or controlling speed or following distance, are assisted by the automation (Levels 1-2), drivers are still expected to be actively engaged and to continuously monitor the automation and driving environment. Adaptive cruise control and lane centering are each examples of Level 1 systems. When combined, as in Tesla’s Autopilot software, they are Level 2. At no point can a Level 2 system, also known as partial driving automation, ever replace the driver.

In contrast, drivers aren’t expected to be actively engaged when using automated systems that monitor the environment in addition to performing some or all parts of the driving task (Levels 3-5). Some of these systems may rely on the driver to intervene in certain situations or in case of emergency (Level 3), while others may not require human input at all while driving under some (Level 4) or all (Level 5) driving conditions.

Most automakers now offer Level 2 driving automation in at least some of their vehicles. Mercedes-Benz announced it would offer Level 3 driving automation to consumers in the United States in the 2024 model year, becoming the first automaker to do so (Mercedes-Benz, 2023).

So far, all of the technology that is currently available to consumers is constrained to specific road and environmental conditions, so drivers will be expected to bridge the gap until full driving automation is developed that can perform all aspects of driving without human input under all conditions (Level 5).

While partial driving automation may be convenient, we don’t know if it improves safety. Recent studies by HLDI have yielded mixed findings about any safety benefits beyond the crash avoidance features typically included on vehicles with such systems (HLDI, 2019; HLDI, 2019; HLDI, 2022).

Much of the technology available in current vehicles, such as adaptive cruise control and lane centering, typically works only on higher-speed roadways where crashes are relatively infrequent. Even if all interstate miles were logged by vehicles driven entirely by automation that did not crash, the maximum overall benefit would be 17% fewer crash deaths and 9% fewer crash injuries than when driven by a human without automation assistance (IIHS, 2016).

Partially automated systems may also have unintended consequences, as drivers become disengaged because the vehicle is handling more of the driving (Reagan et al., 2021). System misuse has already been implicated in fatal crashes (NTSB, 2017; NTSB, 2020; NTSB, 2020). To ensure drivers continue to pay attention while using Level 2 systems, automakers use various driver monitoring strategies, some of which work better than others (Mueller et al., 2021). IIHS began rating the safeguards built in to Level 2 systems, including driver monitoring, escalating attention alerts and fail-safe procedures, in 2024.

Drivers also need to understand the limits of Level 2 systems so that they don’t expect the technology to handle more than it was designed to. A recent IIHS study found that people who regularly use these systems often have a false sense of security about what the technology can do (Mueller et al., 2022). Other studies have shown that over-trusting these systems can prevent people from intervening even when they see a hazardous situation occurring in front of them (Schneider et al., 2022; Victor et al., 2018).

It is often assumed that higher levels of driving automation will come with improved safety, but that is not a guarantee (Mueller et al., 2020). The potential injury and fatality reductions associated with full self-driving technology will ultimately depend on whether the technology is designed to avoid crashes or to prioritize consumer preferences, even if doing so comes at the cost of safety — for example, by exceeding the speed limit.

Regulatory frameworks for testing and deploying self-driving cars are being developed in the United States and other countries. The U.S. Department of Transportation has issued guidance to help state lawmakers address testing and deployment of automated vehicle technology and to encourage a consistent legislative approach nationwide.

In 2011, Nevada became the first state to enact legislation specifically permitting research and testing of vehicles with partial and full autonomy on public roads. Since then, most states have passed legislation or issued executive orders addressing driving automation. Some of the laws only authorize a study, define key terms or authorize funding, while others permit testing on public roads or authorize full deployment. In some states, testing on public roads has been allowed without any specific legislation or regulatory action.

Autonomous vehicle laws by state, in detail

Vehicles that drive themselves are only available right now as part of taxi and ride-sharing services and are not available to consumers for purchase. In August 2016, the world’s first taxi service featuring high levels of automation debuted in Singapore (Watts, 2016). In 2016, Uber began allowing select customers in Pittsburgh to hail a vehicle that drives itself with a human supervisor, marking the first time the public could ride in this type of vehicle in the United States (Hawkins, 2017). On October 8, 2020, Waymo began its driverless ride-hailing service in the Phoenix metro area (Waymo, 2020) and, along with other companies (e.g., Cruise and Nuro), has since expanded its service to the San Francisco area. 

Even though these vehicles do not have a driver behind the wheel, they are still remotely monitored by trained employees who can intervene when a vehicle encounters conditions that are difficult for it to understand (Waymo, 2020).

Regardless of when vehicles equipped with high levels of automation become available for purchase, it will be decades before most vehicles on the road drive themselves. It takes a long time for new vehicle features to penetrate the vehicle fleet. For example, electronic stability control was introduced in the United States in 1995 models, but it was not until more than 20 years later in the 2017 model year that it became standard on all new vehicle models (HLDI, 2023). Since most people don’t drive brand new vehicles, it won’t be before the 2030s that 95% of the vehicles on the road have ESC.

More recent crash avoidance technologies like front crash prevention and lane departure warning are expected to be nearly universal only after 2040 (HLDI, 2023). It will be even longer before most registered vehicles in the U.S. are equipped with Level 2 automation.

Vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) communications, collectively known as connected vehicle technology (or V2X), are safety systems in which vehicles and roadway infrastructure communicate with each other over a wireless network.

With V2V communication, vehicles transmit information regarding their actions to other vehicles. For example, in a long chain of vehicles, if the lead vehicle suddenly brakes, this information will be transmitted to every other vehicle in the chain so that the other vehicles are notified. It also could be possible for the trailing vehicles to automatically begin braking when the lead vehicle’s signal is received.

With V2I communication, cars receive from and transmit information to roadway infrastructure. For example, highway systems could monitor vehicle location within a lane. If a vehicle is detected to be drifting out of a lane, the system could alert neighboring vehicles. In urban environments, traffic signals can alert vehicles of an impending light change so drivers can prepare to stop or the vehicle can automatically slow down if the driver does not.

Two V2V functions, intersection movement assist and left turn assist, warn drivers or intervene on their behalf when they are at risk of crashing with another vehicle when entering or turning left at an intersection. NHTSA estimates that together these functions could potentially prevent up to nearly 600,000 crashes and about 1,300 fatalities annually when fully deployed through the light vehicle fleet (Office of the Federal Register, 2017). Recent IIHS research shows they could be of particular benefit for older drivers, who are more often involved in intersection crashes (Cox et al., 2022).

However, there are some barriers to wide adoption, such as the FCC’s regulatory restrictions that limit the spectrum available for the technology.