BLUESKEYE AI completes Zenzic CAM Scale-up data collection to measure fatigue in vehicles

Too many road accidents are caused by people falling asleep at the wheel. In the US alone in a single year 91,000 police-reported crashes involved drowsy drivers, with an estimated 50,000 people injured and nearly 800 deaths.

Many of these accidents are preventable by the use of technology, in particular by the installation of driver monitoring systems. This is technology that, in theory, measures all your expressive behaviour to infer your mental and physical fatigue levels. As it turns out, it’s not that advanced. Yet. But BLUESKEYE AI is about to change that!

Let me paint the current scenario for you:

Since July 2024, all new cars in the EU must have a driver facing camera to measure driver drowsiness and attention and warn them when needed (so-called DDAW systems). The Vehicle General Safety Requirements (GSR) is EU legislation that sits alongside Euro-NCAP safety standards and is a requirement rather than a safety rating. Euro-NCAP provides a safety rating scale rather than a legal minimum, so the benefit is that car makers are rewarded for making cars safer than the absolute minimum. They’ve been around for years with a well established testing ecosystem and detailed test protocols. The final protocol for driver drowsiness and attention was released very recently, in March 2025, with an implementation date of January 2026.

That’s a great start to improving road safety. To begin with, all new vehicles placed on the market in the EU will now have driver-facing cameras, which can be used to determine that a person’s eyes are closed, and roughly what they are looking at. And that’s exactly the approach taken by the GSR and Euro-NCAP. Specifically, Euro-NCAP defines the different sleeping-at-the wheel related driver impairment states as:

• Drowsiness - A driver reaching a Karolinska Sleepiness Scale (KSS) level ≥7, or an equivalent metric appropriate to assess risky levels of drowsiness.

• Microsleep - A driver is considered to be undergoing a microsleep event when displaying a short duration eye closure (1-2 seconds).

• Sleep - A driver is deemed asleep when displaying a continued eye closure ≥3 seconds.

• Unresponsive driver - A driver may be classified as unresponsive when their gaze does not return to the forward road view within 3 seconds after a distraction warning being issued or when the eyes have been closed for ≥ 6 seconds.

This only solves part of the problem.

Improving safety is not only about acting when the driver has already fallen asleep. Once someone is asleep, the only thing you can do is to bring the car to a safe stop, in the absence of level 5 driving automation. It would be much better to prevent impaired driving, which should start well before people fall asleep by noticing people are starting to get fatigued and predicting by when they would actually fall asleep. It would then warn the driver so they can take a break, rest, and recuperate rather than executing an emergency stop .

Unfortunately no currently available driver monitoring system really measures fatigue, only eye closure, and they aren’t able to predict future fatigue levels. This is too simplistic and does not enable the driver to be properly warned of onset of fatigue.

BLUESKEYE AI models mental state over time and has clinical grade technology to predict fatigue and sleepiness ahead of time.

To prove our technology’s capabilities, we collected data in an actual car on an actual test track.

As part of a UK government funded project, managed by Zenzic, we had 20 participants drive around in circles for four hours, on the outer loop of Horiba Mira’s Warwickshire test track. To ensure they were safe, there was a safety driver sitting next to them.

As if driving around in circles wasn’t bad enough, after two hours driving we turned up the heat, turned off the radio, and the safety driver would stop engaging in chit-chat. Needless to say, we got some very drowsy participants! At regular intervals, the participants would verbally complete the KSS scale until they were so tired that the safety driver instructed them to stop the car.

Here’s a 6 minute video made by Zenzic that gives a nice visual representation of the data collection and explains the reasoning behind it.

So what’s next in the world of driver safety monitoring?

Whilst it’s clear what is required for the 2026 implementation date of Euro-NCAP, much less is known about the 2029 implementation date. All we have to go by is a rather short 2030 vision paper which is starting to feel rather outdated. Regardless of this, the external validation of our drowsiness and fatigue measurement system sets us up perfectly for delivering on whatever fatigue related requirements Euro NCAP 2029 will throw at the automotive industry!

So, if you’re interested in learning more about BLUESKEYE AI’s advanced driver monitoring capabilities, then get in touch with us today!

PS: I have some interesting thoughts about what the 2029 implementation may have in store for us in other areas of driver impairment monitoring, but that’s for another day.