When a leading vehicle suddenly brakes or an oncoming car unexpectedly enters your lane, you have only fractions of a second to decide whether to brake, swerve, or both. "Existing models typically describe only part of this process, such as reaction time or steering behavior," says Arkady Zgonnikov, Assistant Professor at Delft University of Technology (The Netherlands). "Our new model brings all these components together." The model integrates perception, decision-making and execution into a single coherent framework. As a result, it can detect when a situation becomes dangerous, predict how the traffic situation is likely to evolve, and simultaneously determine the most effective avoidance strategy.
To test the model’s validity, the researchers compared it with human behavior in three hazardous traffic scenarios: a leading vehicle braking suddenly, an oncoming vehicle entering the lane unexpectedly, and a car failing to yield. The model was given exactly the same information as human drivers. "The model showed realistic braking reaction times and made similar choices between braking and steering," says Zgonnikov. Moreover, it incorporates human limitations, ensuring that the resulting behavior remains recognizably human.
The researchers see important applications for both the development and evaluation of autonomous vehicles. "It can help address whether autonomous vehicles are safer than human drivers, a key question in regulation," says Zgonnikov. "At the same time, it becomes possible to formulate clear and measurable requirements for manufacturers." According to Mauricio Peña, Chief Safety Officer at Waymo, the model can help the sector to "move towards a shared, scientifically grounded approach to assessing collision avoidance."
Nature Communications
Computational simulation/modeling
Not applicable
Active inference as a model of collision avoidance behavior in human drivers
10-Jun-2026