When the auto-driving systems were launched, the automakers coined these systems as an upgrade of the vehicle’s safety components. In practice, the reality is different, and many refer to automatic driving systems as if they perform fully autonomous driving. Although these systems include measures designed to confirm the driver’s alertness, a study published earlier this year finds that these systems are not effective enough
Stephen Hendrickson from California thought he had hit the jackpot. At the end of each workday, he used to get into his Tesla 3, turn on the autopilot, lay back, and rest or scroll the social networks.
He was so enthusiastic about his revelation that he began documenting his “driving” and uploading videos of himself on YouTube and Tic Toc. On May 5, 2021, Hendrickson’s show ended with a bang when his Tesla 3 collided violently with a truck, and he was killed on the spot. Hendrickson’s case is particularly extreme and colorful, but sadly he’s not alone. There have been dozens of reports of accidents involving Tesla 3 and other cars equipped with automatic driving systems in the last two years.
Automatic driving is not autonomous driving
Part of the problem is that there is conceptual confusion in public. Automatic driving systems, such as the one in Tesla 3 or Cadillac Escalade, are known as Active Driving Assistance Systems or ADAS for short. These systems show an impressive display of technology light years ahead of the Cruise Control systems we have met in the past. Some of them certainly create a sense of autonomous driving, which is precisely the problem.
The concept of autonomous driving refers to driverless cars that can carry passengers from point to point with zero errors. Some of the developments in this area, including the latest update of Tesla’s system, are close to reaching this goal, but “close” is not enough on the road. Although the new systems know how to deal with many descriptions and driving scenarios, at the end of the day, the road is the “kingdom of uncertainty,” and there are situations where even driving in ADAS mode requires quick driver intervention. Too many drivers who use these systems are trapped in the illusion of “autonomous driving” and allow themselves to lower their alertness level and dive into the cell phone. Unfortunately, as in the case of Stephen Hendrickson, sometimes the results are disastrous.
The built-in distraction prevention systems aren’t efficient
This is the place to point out that although some manufacturers contribute to the drivers’ sense that the system allows for fully autonomous driving, their official position is that the autopilot cannot replace the human driver entirely and should not remove from him the responsibility for what happens on the road.
Therefore, all systems include measures to verify the driver’s alertness and prevent distractions. Currently, the two technologies adopted by the manufacturers to monitor the driver are cameras that detect whether he is looking at the road and sensors on the steering wheel that detect the driver’s condition. If these measures detect a lack of alertness and attention on the driver’s part, they will respond with a series of warnings designed to divert his attention.
Despite these measures, we have seen an increase in the number of accidents involving cars traveling in ADAS mode in the last year. An experiment was recently conducted to test their effectiveness; as part of the experiment, four models of vehicles with an ADAS system that allows for a “like-autonomous” driving experience (or, as some manufacturers problematically define it as “semi-autonomous driving”) were selected.
Two of the tested models, the Tesla 3 and Honda Santa Fe, are equipped with steering wheel sensors to verify driver alertness. The other two models, the Cadillac Escalade and Subaru Forester are fitted with camera-based systems.
Three tests reveal the limitations
The experiment, designed to test the effectiveness of systems in cases of distractions, included three different tests, which together provided a picture that revealed the limitations of existing systems. The experiments took place in a fixed trajectory and period. A 10-minute drive at a speed of 100 km / h on a closed toll road.
In the outline of the first experiment, drivers were asked to raise their hands but not touch the steering wheel and turn their heads down throughout the ride; In the outline of the second experiment, drivers were asked not to touch the steering wheel but to look directly at the dashboard; Whereas in the third experiment the drivers were asked not to touch the steering wheel but to try to deceive the system, using hand movements and head movements from time to time.
The findings in all the descriptions were problematic. It turned out that steering wheel sensor-based systems are very slow in detecting the driver’s lack of alertness, with some tests out of a 10-minute drive (which, as part of the experiment, is defined as “distracted driving”) for more than five and a half minutes without alerting.
In camera-based systems, the situation was less severe. Still, it was also very far from addressing the problem. Even in camera-based systems, it was possible to drive distractingly for two and a quarter minute without receiving any warning. The experiment also revealed that drivers could perform manipulations on the system that would cause it to issue fewer alerts. Although the recommendation was to prefer camera-based designs in light of the experiment results, the general feeling is that this solution is partial and far from satisfactory.
SaverOne may break the paradigm
The solution to making ADAS systems more secure may be via SaverOne’s technology in the future. Even during “autonomous driving,” most of the distractions that steal drivers’ focus from what is happening on the road are cellular distractions.
Presumably, anyone who is convinced that his car can take him home safely, without his involvement, will allow himself to read, play or write on the cell phone. Therefore, integrating SaverOne technology as part of the interface of ADAS systems may make the difference and safety package of these systems much more efficient. And that will bring us even closer towards fully autonomous driving, which is also really safe.
