Simulation Measuring Visual-Cognitive Distraction in Real Time

Related Vendor

FAULHABER Drive Systems

Researchers in Milan have developed a real-time metric to measure visual and cognitive distraction while driving. The tests are conducted in a simulator.

Distraction while driving is one of the most common causes of accidents worldwide. As more driver-assistance systems and semi-automated features are installed, the number of secondary activities that drivers engage in while driving is increasing. Legislators are responding: UNECE Regulation No. 171 requires Driver Control Assistance Systems (DCAS) to monitor driver attention; and Euro NCAP’s Roadmap 2030 calls for systems capable of detecting cognitive distraction.

Existing driver monitoring systems (DMS) typically rely on camera-based eye-tracking and thus primarily detect visual distractions. If a secondary task also requires mental processing—such as reading and calculating values on a display—experts refer to this as visual-cognitive distraction. It is considered particularly difficult to quantify because it cannot be determined solely by eye-tracking data.

A research team at the Politecnico di Milano has developed a new metric for this purpose: VICODEV, which stands for “VIsual-COgnitive Distraction through Eye-tracking and Vehicular measurements.” The study, published in the journal IEEE Open Journal of Intelligent Transportation Systems, combines vehicle and eye-tracking data into a single, continuous distraction index—derived from experiments conducted in a dynamic driving simulator.

The Test Setup

The “DiM400” driving simulator was developed by the manufacturer VI Grade in collaboration with the Politecnico di Milano. The university has been operating it in its in-house “DriSMi” laboratory since 2021. The two-stage motion system—a cable-guided platform for horizontal movement (4 × 4 meters, ±60 degrees of yaw) and a Hexalift with six degrees of freedom up to 30 hertz—is supplemented by eight shakers for high-frequency vibrations up to 200 Hz.

A 270-degree projection, active seat belts, an active brake pedal, and a cockpit based on a real vehicle frame provide a realistic driving experience; system latency is less than 20 milliseconds.

In the simulator, 35 inexperienced drivers—21 men, 14 women, aged 19 to 30—drove a 4.1-kilometer (2.5-mile) route modeled after a neighborhood in Wolfsburg (Germany), with 107 simulated vehicles. To keep the actual driving task consistent, the participants continuously followed a vehicle ahead of them in the left lane. This was intended to ensure that any differences between the test phases were attributable solely to distraction and not to changing traffic conditions.

Tests Too Dangerous for the Road

An experiment of this kind would be unacceptable due to the deliberate and repeated distractions encountered in real-world traffic. The simulator, on the other hand, made it possible to present all 35 participants with exactly the same distraction protocol—repeated five times—under identical road, traffic, and lighting conditions—a level of control and repeatability that real-world test drives cannot provide. At the same time, driving behavior, eye movements, and physiological signals had to be recorded in precise synchronization so that they could later be assigned to the individual driving phases.

Distraction Using the Clock Task

The established “Clock Task” was used as a targeted distraction. Six clocks appeared on a tablet mounted in the dashboard; the drivers had to add up the times displayed and call them out aloud. The color, font, and size of the clocks varied to increase the visual and cognitive load. Five such 45-second distraction phases alternated with equally long phases of normal driving.

Throughout the entire drive, the system synchronously recorded data from multiple sources: An instrumented steering wheel developed at the Politecnico measured grip force, while eye-tracking glasses tracked eye movements; an electrocardiogram and a 65-channel electroencephalogram were also running simultaneously. Because the simulator’s virtual CAN network runs in a real-time database, all signals could be precisely assigned to specific driving phases after the fact.

From 21 to 5 metrics

In total, the researchers recorded 21 vehicle-related, behavioral, and physiological metrics. Twelve of these showed statistically significant differences between normal and distracted driving. For the Vicodev Index, the team selected five of these parameters that can be measured in real time using production vehicle sensors: the standard deviation of lateral deviation from the center of the lane (SDLP), the standard deviation of steering wheel grip force, the proportion of time spent looking away from the road, the rate of rapid eye movements, and the average duration of slow eye movements.

A multivariate analysis of variance and a canonical discriminant analysis determined the extent to which each variable contributed to the linearly combined index—SDLP and the duration of slow eye movements proved to be the most influential factors.

Validation

In a cross-validation across the five repetitions, Vicodev achieved an average hit rate of 99.47 percent and an accuracy of 95.2 percent. By comparison: Lane deviation alone achieved an accuracy of only about 61 percent, while time spent looking away from the road alone achieved 86.5 percent.

A second validation, in which one test subject at a time was completely excluded from the training dataset, yielded a median AUC value— “area under the curve”—of 82.9 percent and confirmed that the index identifies consistent patterns across different drivers rather than merely reflecting the individual characteristics of specific test subjects.

Conclusion and Outlook

Physiological signals such as EEG and ECG were deliberately excluded from the final index, as they cannot currently be reliably measured in production vehicles. However, the study shows that heart rate variability, for example, also reacts significantly to induced distraction—a potential starting point for future systems supplemented by wearables.

The authors also emphasize that their findings currently apply only to the group of young, inexperienced drivers studied and to the clock task used. Further studies involving other age groups and different tasks are planned.

For the automotive industry, this work provides a methodological building block toward the cognitive distraction detection systems required by Euro NCAP starting in 2030 —developed and validated in a simulation environment that provides the necessary controlled, repeatable, and safe test conditions.

Lorenzo Uccello, M.Sc., Ph.D. candidate in the Department of Mechanical Engineering at the Politecnico di Milano. His research focuses on driver assistance systems, automated driving, and human-machine interaction.
Corrado La Russa, M.Sc. Graduate of the mechanical engineering program with a focus on automotive engineering and motorsports at the Politecnico di Milano. As part of his master’s thesis at the DriSMi driving simulator, he investigated the monitoring of driver status using an instrumented steering wheel and eye tracking.
Prof. Dr. Massimiliano Gobbi, Full Professor of Mechanical Engineering at the Politecnico di Milano. His research interests include automotive engineering, the optimization of complex systems, and mechatronics.
Prof. Dr. Gianpiero Mastinu, Full Professor at the Politecnico di Milano specializing in road and rail vehicles. He is a co-founder of the university’s research laboratory for dynamic driving simulation and focuses on vehicle development, vehicle dynamics, and experimental vehicle testing.

Battery Development and Simulation

5 Approaches to Optimized Battery Designs