Making a Car that Won't Let You Sleep
Algorithms, sensors—there's There's even talk of using biometrics and wearables, data derived straight from a driver's body, to determine if a person is too tired to drive.
Image: Brad Greenlee/Flickr
Shortly before 8 AM on a July morning in 2004, Allan Michael Kasprick drove his Dodge Ram pickup truck into the opposite lane, crushing an oncoming van.
The van's driver and her daughter were instantly killed, while a third occupant, the daughter's friend, died shortly after. Kasprick was convicted on three counts of dangerous driving causing death and two counts of dangerous driving causing bodily harm.
"What is unique about this case is that it involves none of the contributing factors usually encountered in a dangerous driving case," Justice George Baynton wrote in his decision the following year. "The evidence is clear that the accused had not been drinking, had not been taking drugs, had not been speeding or stunting and had not been operating his vehicle in a reckless fashion."
Rather, "the evidence is also clear that the ability of the accused to drive his vehicle was impaired by sleep deprivation and drowsiness."
There have been no shortage of anecdotes and rulings like Baynton's over the years, and for the better part of a decade, car manufacturers have been trying to build cars that don't let drivers fall asleep at the wheel. There are algorithms that can detect when a car begins to swerve between lanes and sensors that watch for drooping eyes. There's even talk of using biometrics and wearables, data derived straight from a driver's body, to determine if a person is too tired to drive.
The reason is pretty simple: Drowsy driving caused "about 72,000 crashes, 800 fatalities and 44,000 injuries" in 2013, according to the National Highway Traffic Safety Administration's most recent data. And until self-driving cars are able to ferry our sleepy selves from the office or bar to the house and back without putting us being behind the wheel, stop-gap technologies with names like "Attention Assist" are our best bets at bringing those numbers down.
There are a few different methods that car manufacturers and researchers have used to detect drowsiness in drivers. One of the cheapest, and most common, leverages a vehicle's existing assistive driving systems, according to Mark Boyadjis, a senior analyst of automotive infotainment systems and human-machine interfaces at market intelligence company IHS."These systems help you to stay in your lane, they help you to manage distance, and to understand where objects and vehicles are around you," Boyadjis explained. "So while their primary function is to act as a bit of a safety net, they can also determine, after a certain period of time [...] how often you're swaying left and right on a roadway, or your following distance, or your acceleration and braking at the time. And collecting all this data together, they can kind of create a virtual profile of the driver in real time, and they can take into account all these different systems to determine whether you are drowsy or not."
Ford, for example, has its Driver Alert System. Nissan has Driver Attention Alert. Infiniti has a Lane Departure System. Subaru has EyeSight. Acura, Volkswagen, Daimler, Mazda, Tesla, Audi and more all have similar features. Some of these companies have been refining their features for nearly a decade, but they all essentially work the same way, using algorithms to compare a driver's driving style during the first few minutes of a trip with their performance as time goes on.
If a driver is suspected of dozing off, an alert goes off, warning the driver to get off the road and take a break.
More recently, the declining cost of cameras and infrared sensors has resulted in some higher-end systems that can go a step further, and identify changes to a driver's eyes—tell-tale signs that might indicate drowsiness or inattentive driving.
"The more expensive solution, is to literally put two infrared sensors, or an array—generally you only need two, but sometimes you might use three or four—on the top of the steering column," Boyadjis said, "pointed directly at the driver's face, for the purposes of watching your eyes, watching your face, watching your head movements, your focus on driving versus your focus on other things, or your lack of energy and lack of focus altogether."
Boyadjis suggested that it's only a matter of time before automakers begin to rely on data from wearables such as watches and Fitbits
At present, however, neither driver attentiveness or infrared sensing systems are typically sold as standard features according to both Boyadjis and University of Waterloo computer science professor Jo Atlee, but rather, available as upgrades to base models or on higher-end vehicles, which has inherently limited their adoption.
According to Atlee, who co-leads an automotive software research group that has worked with General Motors Canada on analyzing assisted driving technologies, such systems could one day be mandated as standard features by transportation regulators if there are clear benefits for all drivers.
"You cannot write software that is going to work 100 percent of the time," Atlee explained. "But simply the fact that it was demonstrated that cars with those features were much safer than cars without those features, that was enough for the government organizations and regulators to require the features in all vehicles. And so I would expect the same thing to happen in many of these other active safety features."
Perhaps the most interesting direction for such technology is in gathering more detailed biometric information directly from drivers—going beyond mere facial recognition to the measurement of breathing rate, blood pressure, motion activity and more.
For example, in 2011, researchers at Germany's Technische Universitaet Muenchen in collaboration with BMW installed sensors in a steering wheel that could measure "vital signs such as heart rate, skin conductance and oxygen saturation in the blood via simple sensors in the steering wheel," according to a release.
More recently, Spain's Biomechanics Institute of Valencia published a video in 2014 detailing research on the measurement of heart and respiration rate using sensors embedded in a car's seat. The system was called HARKEN, or "heart and respiration in-car embedded nonintrusive sensors."And this past December, Toyota's US spokesman John Hanson told Automotive News that, "Toyota is actively researching a variety of methods to monitor a driver's physical and mental state, such as eye monitoring, body temperature, heart rate and blood pressure, via a wide array of sensors, including wearables."
Boyadjis suggested that it's only a matter of time before automakers take their connected car plans a step further, and begin to rely on data gathered from wearables such as watches and Fitbits, which can track biometric information that could be useful in measuring a driver's state before they even enter the car.
"A car maker could spend a bunch of money embedding sensors into steering wheels and seats, or they could spend a lot less money integrating something maybe you're going to be wearing already," suggested Boyadjis.
Of course, there will be no shortage of privacy concerns to address as our cars effectively turn into mobile diagnostic centers on wheels. "Is it okay for the self-aware vehicle that you own to kind of spy on you?" Gartner automotive analyst Thilo Koslowski mused in a Fast Company article from 2014 on detecting inebriated drivers in-car. "And that's where I think a lot of the car manufacturers still are a little bit careful that they don't want to overdo this, because of that expectation that people have in their car to do whatever they want."
"These sensors could really add some value adds for safety, or an opportunity for integrating that personal safety in the car, [but] you need to communicate it in a proper fashion," Boyadjis offered. "I don't want to say it's all just a PR story, but I that's an important element, because adding sensors without really describing how they're going to be beneficial […] the customer's not going to be interested."
Not to mention, many of these systems may be rendered moot before long.
"The future really is the self-driving, where the software would have complete control over the driving experience," Atlee said. "The features we've been talking about so far are baby steps."
You'll Sleep When You're Dead is Motherboard's exploration of the future of sleep. Read more stories.