A recent study suggests that it's possible to detect a driver's level of concentration and attention by monitoring their eye movements. This method could prove very useful in semi-autonomous cars, especially when the vehicle signals the driver to retake control.
In some countries, semi-autonomous cars are already permitted. These are vehicles that offer an "autopilot" mode, allowing the driver to release the steering wheel in scenarios like traffic jams. However, the vehicle may at any point request the driver to regain control, particularly when traffic flow improves, and a certain speed is reached. The driver must then quickly take over for total control of the vehicle.
Still from Stanley Kubrick's "2001: A Space Odyssey"
Scientists from University College London (UCL) sought to determine whether it was possible to detect a person's attentiveness: Is the individual focused or too engrossed in another task to be able to respond quickly to such a signal? In their research, they exposed 42 participants to experiments recreating this scenario.
Specifically, the participants were engaged in an activity on a screen, with two levels of difficulty. The first level involved identifying simple shapes (for example, finding the letter "L" among a series of "T" letters); the second required spotting more precise combinations of shapes and colors, demanding greater concentration. Participants were instructed to immediately stop their activity and press a button as soon as they heard an auditory signal.
The analysis revealed that the more complex the task and the more it engaged the participant's attention, the longer was their reaction time to the auditory signal. The researchers then developed a method to identify the participant's level of attention, achieving success by analyzing their eye movements. They found that the shorter the eye movements and the longer the fixations, the more concentrated the person was on a demanding task, and consequently, the less reactive they were to auditory signals.
The researchers subsequently trained an AI to automatically detect a person's attention level. They believe that, while the AI still needs training with a more substantial data volume, the early results are promising. Eventually, this process could reliably identify if a driver is capable of quickly reacting to an alert signal to retake control of the vehicle or if they are too focused on another activity to be responsive.
Article Author: Cédric DEPOND