Tesla, Honda and many other automakers that offer advanced driver-assistance systems have been stung by a phenomenon known as "phantom braking," which reports say causes vehicles to suddenly slow -- or stop -- in response to what the system mistakenly perceives as a threat.
Phantom braking -- also known as unintended braking -- is caused by erroneous sensor perception.
In February, it was reported that Tesla was facing a probe in U.S. after authorities received more than 354 complaints for unexpected brake activation in the 2021-22 Model 3 and Model Y.
In August, a California owner of a Tesla Model 3 sued the electric vehicle maker in a proposed class action related to phantom braking, calling it a "frightening and dangerous nightmare," according to the lawsuit.
Honda is under investigation by the U.S. National Highway Traffic Safety Administration (NHTSA) for a phantom braking issue affecting more than 1.7 million vehicles. NHTSA has received more than 270 complaints that allege braking incidents "occurring with nothing obstructing the vehicle's path of travel," with six people alleging they were involved in a collision with minor injuries.
"The automatic break [sic] warning came on with no one in front of me and the car came to a halting stop. Thankfully the car behind me was paying attention otherwise i [sic] would have been rear ended," a 2018 Honda Civic owner wrote to NHTSA in March.
Gartner analyst Pedro Pacheco is unaware of any official statistics on how often the problem occurs, the consequences range from an annoying and scary to potentially causing a rear-end collision.
While the actions against Tesla and Honda put their driver-assistant technology under scrutiny, they are not alone -- and the problem is not new.
A 2017 Mercedes E-Class owner wrote in a complaint to NHTSA in August 2021: “The vehicle independently came to an abrupt stop.” While that same month the agency got this from the owner of a Volvo S60: “The automatic emergency braking engaged while we were travelling between 60-70 mph … There was no apparent reason for the braking because there was no vehicle or other obstacle to trigger the braking."
NHTSA also started an investigation of "False Automatic Emergency Braking" in the Nissan Rogue in September 2019. Nissan told the agency that it has received 750 vehicle complaints "relating to false positive activation" of the automatic emergency braking system. Of these complaints, 12 said a collision resulted from the activation of the system, with four people alleging there were injured.
Never 100% performance
"I have seen how these systems are being developed, and you can never make 100 percent sure these systems are going to deliver full performance across a wide variety of situations," said Adriano Palao, technical manager of ADAS and automated driving for car testing agency EuroNCAP
Eliminating phantom braking is especially difficult in Europe “where we have a wide variety of infrastructure across countries," Palao said.
Gartner's Pacheco said a solution is to "use a plethora of sensors such as radar, plus lidar, plus cameras, plus advanced machine learning that has been trained to deal with this particular situation.” However, he believes the problems start when automakers want to cut on the number and performance of sensors or how well the system is trained to address a particular situation.
"This is, obviously, due to cost considerations," he adds.
Rebuilding loss of trust
EuroNCAP's Palao said phantom braking is "a huge problem" because it undermines confidence in what is designed to be a potentially life-saving system.
"At the moment, this system does not deliver what it's supposed to deliver, and the driver does not trust in the vehicle they have just purchased," he said. "Then the safety benefit is just gone because the driver will no longer trust the systems, and they will start switching them off."
He adds that although one could argue that the cases of reported phantom braking will increase because there are more ADAS-equipped vehicles on the road, the technology is improving in "giant steps" every year and the systems on the market are delivering performance in situations that 10 years ago would have been unthinkable.
To continue this improvement, EuroNCAP will devise ways to address the robustness of ADAS to determine which brands offer the solutions best equipped to handle a wide range of challenging real world scenarios, Palao says.
Looming challenge
A looming challenge to solving the phantom braking problem is the roll out of ADAS to lower-segment models because there is pressure to reduce the cost of the system.
"If it sees something that's not really there, can you trust the system to operate your car for you, totally or even partially?" Gartner's Pacheco said. "This is fundamentally about how automakers juggle three fundamental priorities: ADAS performance, safety margin and cost."
He cautioned that if automakers fail to make their ADAS 100 percent reliable it will stall the move to full-autonomous vehicles.
"In addition, regulators also must play a stronger role – ADAS or AV technology should not be regulated purely on feature definition but its capability envelop must also be defined and tested according to an homologation program," he said.
1 per 1 million km
Manja Greimeier, head of ADAS for German technology supplier Continental, said that data fusion is key when advance driver safety systems.
"We believe that with the addition of lidar, for example, we will be able to support high-end applications and reduce the number of false positives," she said. "We are continuously improving our components, learning from different fusion scenarios, and also learning from having more cars on the road."
An advantage of over the air (OTA) software upgrades is that the vehicle can learn as the driver assistance systems compile real world data that will help it better handle a wider array of situations.
Greimeier said that Continental aims to reduce the frequency of phantom braking occurrences to one for every million kilometers driven. Failing to do reach that level of success undermines confidence in the system, she said.
"That's not something that we can consider a good system," she said. "If that happens, I won't feel comfortable in the car. It's our ambition to reduce the number of false alarms."
Lidar's role
Bernd Reichert, senior vice president of automotive for AEye, whose full-stack solution combines lidar, camera, and radar technology in Continental's integrated automated driving (AD) platform for passenger and commercial vehicles, says he is "fully convinced" lidar will help reduce the number of phantom braking incidents.
"The key is a comprehensive sensor set, which acts like a force multiplier and can even exponentially increase the intelligence of a part of the software," he said.
He said that reducing phantom braking not only improve safety, it also boost the overall user experience.
"In the end, it stands and falls with the trust you have in the system," he said.
Both Greimeier and Reichert admit they don't know at this point how much money must be spent to adequately resolve the issue of phantom braking.
"From a sensor point of view, it's super important that we try to get intelligent data and that we have good quality of data to help eliminate false positives," Reichert said.
Palao agrees that a robust mix of sensors could help minimize phantom braking.
"The OEMs all know as much as I know, that a system that only relies on a particular kind of sensor will at some point have a coverage gap," he said. "A high degree of coverage is accomplished by fusing different sensors, where the gap of one sensor is complemented by the strength of the other. This is the reality."
Reality, of course, is extremely nuanced and complex, but Palao remains positive.
"Every single variation you can encounter, even if you do the same drive every day, the light conditions can be different, the temperature can be different," he noted. "But I believe as an engineer that things can change for the better, and we are going to learn from our mistakes."
