Photo: Tesla | Edited
Tesla is about to improve its camera-based driver monitoring system by adding the ability to detect drowsiness. The camera will be able to track additional parameters even when the car is not on Autopilot, significantly boosting safety.
Many vehicles come with a drowsiness detection system in standard. The more basic systems analyze the steering wheel movements and input from the lane-centering system and other sensors to detect how tired the driver is. The best systems on the market also use a cabin monitoring camera (or cameras) to track the driver's eyes and head position, offering more accurate recommendations. Tesla falls somehow in between with a controversial approach to driver monitoring.
Tesla Model 3 and Model Y launched with a cabin camera, but in the case of the Model S and Model X, a camera was added only with the 2021 refresh. Still, even for the newer models, Tesla started using the camera for driving monitoring in May 2021. Until then, the Autopilot system would require drivers to apply pressure on the steering wheel for the system to keep working. Some creative people learned that attaching a soda can or something to the steering wheel would trick the system into thinking the driver is in control.
It was one of the reasons why Tesla Autopilot consistently ranked behind others in terms of safety, despite superior capabilities. Even after Tesla turned to the cabin camera for driving monitoring, it made it easy to fool the system. Some reported that covering the camera or using polarized glasses would force the system to fall back to the old steering wheel check, rendering it useless. In time, Tesla has improved the driver monitoring system to make it harder for people to trick the car. Still, it never offered drowsiness detection or recommendations to drivers based on their behavior.
This is set to change, according to white-hat hacker Green (@greentheonly), who regularly analyzes Tesla software for changes. An upcoming OTA update will significantly boost the camera-based driver monitoring system in Tesla vehicles. Based on his findings, the camera will analyze how many yawns the driver had recently, how often they blinked, and for how long to calculate drowsiness. The system will also analyze lane centering, as well as lane keep assist warnings and corrections to figure out if the driver is fit to drive.
This will work regardless of whether the car is on Autopilot or the driver is in full control. It is unclear what happens when the driver shows signs of drowsiness. Green presumes the sound warnings would go off more often and be more annoying. Hopefully, the active driver assist systems will also become more cautious. Other carmakers offer the option to stop the vehicle when the driver becomes incapacitated (falls asleep or has a medical problem). This would turn the car into an obstacle on the road and increase the risk of rear-ending, so it's not entirely safe either.
For now, the code has just been added to the software, but no change appears in the user interface to reflect these changes. Camera-based driver monitoring works well, but an infrared camera is much better, especially during night drives. Still, as long as covering the camera disables the safety features, it's likely that drivers would be inclined to bypass the new system to get rid of unwanted warnings.
Tesla Model 3 and Model Y launched with a cabin camera, but in the case of the Model S and Model X, a camera was added only with the 2021 refresh. Still, even for the newer models, Tesla started using the camera for driving monitoring in May 2021. Until then, the Autopilot system would require drivers to apply pressure on the steering wheel for the system to keep working. Some creative people learned that attaching a soda can or something to the steering wheel would trick the system into thinking the driver is in control.
It was one of the reasons why Tesla Autopilot consistently ranked behind others in terms of safety, despite superior capabilities. Even after Tesla turned to the cabin camera for driving monitoring, it made it easy to fool the system. Some reported that covering the camera or using polarized glasses would force the system to fall back to the old steering wheel check, rendering it useless. In time, Tesla has improved the driver monitoring system to make it harder for people to trick the car. Still, it never offered drowsiness detection or recommendations to drivers based on their behavior.
This is set to change, according to white-hat hacker Green (@greentheonly), who regularly analyzes Tesla software for changes. An upcoming OTA update will significantly boost the camera-based driver monitoring system in Tesla vehicles. Based on his findings, the camera will analyze how many yawns the driver had recently, how often they blinked, and for how long to calculate drowsiness. The system will also analyze lane centering, as well as lane keep assist warnings and corrections to figure out if the driver is fit to drive.
This will work regardless of whether the car is on Autopilot or the driver is in full control. It is unclear what happens when the driver shows signs of drowsiness. Green presumes the sound warnings would go off more often and be more annoying. Hopefully, the active driver assist systems will also become more cautious. Other carmakers offer the option to stop the vehicle when the driver becomes incapacitated (falls asleep or has a medical problem). This would turn the car into an obstacle on the road and increase the risk of rear-ending, so it's not entirely safe either.
For now, the code has just been added to the software, but no change appears in the user interface to reflect these changes. Camera-based driver monitoring works well, but an infrared camera is much better, especially during night drives. Still, as long as covering the camera disables the safety features, it's likely that drivers would be inclined to bypass the new system to get rid of unwanted warnings.
But it also looks like they are planning to apply this even when not on AP (which is a very right move!) by seeing how well centered the driving is, how many lane keep assist warnings and corrections happened lately.
— green (@greentheonly) May 13, 2023
Google News