At the heart of automatic emergency braking are complex subsystems that must accurately recognize the target, assess risk, create a strategy for the action and deploy it.
It's not an easy set of tasks, and an occasional misperception or misjudgment can trigger sudden braking that could result in a crash.
As automakers grapple with an impending new standard for automatic emergency braking from NHTSA, they must figure out how to meet the mandate without unduly increasing faulty brake activation.
Generally, an automatic emergency braking system perceives information, makes a decision and then initiates an output command, such as a visual, physical or auditory warning, or braking. While this may sound simple, many factors impact system performance. How the vehicle is moving through the environment, traffic conditions and the curve of the roads are funneled into the decision-making process. The system can also observe and act upon weather, lighting and road conditions, and speed and slope.
Complex factors such as tire air pressure, road type and even the driver's mood can impact how the system should respond to an obstacle.
The current "system is much, much more complicated than previously developed" advanced driver assistance systems such as blind-spot warning, said Guoyuan Wu, a researcher at the Center for Environmental Research and Technology at the University of California, Riverside.
"Right now, we try to integrate all the modules: perception, decision-making, planning and also control actuation together, which actually, I would consider as a mini version of autonomous driving," he said.
It's about to get tougher. Automakers must meet a new NHTSA standard by September 2029. It requires automakers to successfully brake automatically for hazards at higher speeds.
There are many factors that can cause inadvertent activation of the system. Perception tools can be either insufficiently accurate or obscured in poor lighting conditions, fog, snow or sandstorms.
Greg Brannon, director of automotive engineering for AAA, said that in a joint study with Michigan Tech Research Institute, researchers found that "if you're going to develop a system that is effective across a variety of different weather, lighting conditions and road conditions, you need all the sensors you can get."
Some sensors are expensive. Lidar, which bounces laser pulses off objects to measure distance more precisely than cameras and has higher resolution than radar, can be hundreds or even thousands of dollars more per sensor depending on the model.
There are 23 models across five automakers that come with lidar sensors, but the Volvo EX90 is the only one that uses lidar to support automatic emergency braking, according to global automotive market research firm SBD Automotive. The remaining 390 models with automatic emergency braking use a combination of camera, ultrasonic and radar sensors.
The industry must "come up with a strategy to fully leverage what's already available and then add in newer technologies, perhaps lidar and other things that will help get it to the new 2029 standard level," Brannon said.
The Alliance for Automotive Innovation surveyed its members and estimated that additional costs to make current systems compliant with the new NHTSA standard range from $200 to $4,200 per vehicle.
The inadequacy of sensors likely causes some of the false activations under recall or investigation by NHTSA, said Michael Brooks, executive director of the Center for Auto Safety.
"A lot of the problems that we've seen in not just automatic emergency braking but also in other technologies including [Tesla's] Autopilot, and some of those driver assistance" features are caused by "overreliance on cameras or sensors that aren't going to ever be perfect at the task."
A particular challenge is pedestrian detection. Current systems can perceive other vehicles with about 95 percent accuracy, but pedestrian and other road-user detection drops sharply to between 60 and 70 percent, though those figures may vary, Wu said.
Inadequate perception can also force the automaker to err on the side of caution when deciding whether to brake, resulting in more faulty activations. Even if sensors are perceiving the environment accurately, automakers still have to calibrate the decision-making system correctly.
Dave Miller, assistant teaching professor of human factors engineering at Tufts University, co-authored a paper, "The Car That Cried Wolf," which is about driver responses to collision avoidance systems with different sensitivities.
He said that tweaking the response of an automatic emergency braking system relies on signal detection theory once the engineer has maxed out the capabilities of the sensors and software.
"You can either reject more false positives or have fewer misses," he said.
His paper also found that erring on the side of a missed signal or failure to brake upon encountering an obstacle can increase driver vigilance. That means automakers should consider how drivers will build a mental model around the performance of the system: If the system generates too many false positives or even performs perfectly, will drivers ignore the warnings or become complacent? If the system has too many misses, will drivers unnecessarily be injured in a crash?
The problem of false activations for automatic emergency braking evokes airbags, another well-known lifesaving technology, said Bryant Walker Smith, a professor at the University of South Carolina with expertise in legal and policy issues for emerging vehicle technologies.
"Airbags save thousands of lives, and they cause injuries even when they themselves are not defective," he said. "So, this is unavoidable. This is the history of technology or even progress: We replace one set of problems with a new set of problems and just really hope our new set, in aggregate, is less than our old."
Send us a letter
Have an opinion about this story? Click here to submit a Letter to the Editor, and we may publish it in print.
