AutoEvolution – Study Confirms That Alienating Drivers From Driving Is Not the Safest Approach
Editors note: the most important point of the article is that some automation is good, while other automation lessens safety:
there will always be truly effective ADAS, such as Lane Keeping Assist (LKA), which reduces crash probability by 19.1%. The second most effective safety driving aid was the Driver Monitoring System (DMS), which lowered accident risks by 14%. Finally, we also have Autonomous Emergency Braking (AEB), with the third-highest crash rate reduction of 10.7%.
See original article by Gustavo Henrique Ruffo at AutoEvolution
It does not matter how well human beings may drive. There will always be someone saying too many people die in traffic – which is true – and trying to offer a magic solution for the problem. That’s why the promise of selling an autonomous car made an obscure manufacturer turn into the world’s most valuable one in market cap. The deal is that removing drivers from driving can actually increase car crashes.
That’s what a study from SWOV (Stichting Wetenschappelijk Onderzoek Verkeersveiligheid, or Foundation for Scientific Research into Road Safety) confirmed on January 29. Published by ScienceDirect, the study revealed that some forms of advanced driver assistance systems (ADAS) make driving more unsafe than if they were not used. They do that because they induce distraction and overreliance on the system.
The study’s main goal was to rethink the taxonomy of ADAS. The researchers argue that the current classification does not help engineers and policymakers address what could really improve safety. To change that, they have classified these systems according to their urgency and level of control, which makes a lot of sense.
Traditional classification takes into consideration either the function (control, monitoring, or information) or how these systems interact with the driver (informing, warning, intervening, or just providing comfort). Using this last taxonomy as the parameter, information systems do not present any level of control or urgency. On the other hand, warning ADAS have a high urgency but no control over the vehicle. Intervening systems have both control and urgency at the highest levels. Comfort-providing ADAS present high control levels but low urgency.
That gives us an idea of what makes intervening systems so effective, as well as what makes comfort-providing ADAS a problem. If the system has a high control level, it should be able to react when something urgent emerges. However, most of them don’t. That was quite a problem for the cruise control (CC) and the adaptive cruise control (ACC). Both keep speeds without requiring the driver to step on the accelerator pedal, although the ACC can also control the distance from the car ahead through radars or Light Detection and Ranging (LiDAR) equipment.
According to the study, the CC may increase the likelihood of a crash by 12%. The ACC is not as bad, but it is not a much safer option: there’s an 8% higher chance of accidents when it is active. And it is not difficult to figure out why.
Photo: AudiWhen any driver uses it, they take their right foot off the accelerator pedal, where it naturally is while they are driving. Costas Lakafossis’s explanation about Tesla crashes on April 20, 2023, is quite suitable for this as well. Just for context, the Greek engineer and crash investigator was discussing his petition to the National Highway Traffic Safety Administration (NHTSA) after several wrecks that owners said were sudden unintended acceleration (SUA) episodes.
According to Lakafossis, the way Tesla vehicles work breaks what neuroscience classifies as closed-loop control, which depends on feedback to work properly. In driving, you need input from the road, the pedals, the steering wheel, and the surroundings, among others. If you take your feet from where they were supposed to be, you may hit the wrong pedal and swear it was the right one. You will only remove your feet from the pedals or your hands from the steering wheel if you trust the car will take care of everything for you.
Believing any vehicle is autonomous has already led to several deaths. Sadly, the researchers did not investigate any of the autonomous driving systems currently sold, which present Level 2 or Level 3 in SAE’s autonomy scale. Although they have the excuse of being safer than humans, they are really meant to provide comfort while driving, like ACC or CC. In July 2023, I wrote that people should not believe in autonomous driving tech propaganda.
Photo: seansinha/Reddit/Kyle Vogt/edited by autoevolutionIn that instance, Cruise tried to convince everyone that “humans are terrible drivers” with an advertisement campaign. Less than two years later, the startup kicked the bucket in a rather shameful way: after one of its cars ran over and dragged a pedestrian for around 20 feet (6 meters) and failed to report it properly. NHTSA made the company pay a $1.5 million fine for that. It also had to recall 1,200 cars before it went bust.
In October 2024, I insisted on this topic, arguing that autonomous tech was a celebration of human ineptitude when we should focus on training to make drivers safer, even if eventually fallible. For these cases, there will always be truly effective ADAS, such as Lane Keeping Assist (LKA), which reduces crash probability by 19.1%. The second most effective safety driving aid was the Driver Monitoring System (DMS), which lowered accident risks by 14%. Finally, we also have Autonomous Emergency Braking (AEB), with the third-highest crash rate reduction of 10.7%.
The SWOV study is a breath of fresh air in automotive safety discussions, bringing objective elements to the table and helping avoid all the nonsense we hear about human beings being terrible at driving. Some of us can surely be bad at the steering wheel, but the solution is not to put these folks in cars with autonomous tech. It is to prevent them from driving or to train them better so they will not be a risk to other road users. Better drivers, helped by more effective safety aids, are what will bring down traffic casualties.
See original article by Gustavo Henrique Ruffo at AutoEvolution
