Unveiling the dynamics and decision-making of dyadic head-on pedestrian collision avoidance: An empirical study

Pedestrian navigation in crowded environments such as shopping malls, streets, and transport hubs relies critically on the ability to safely and efficiently avoid oncoming individuals. A deeper understanding of the underlying mechanisms is crucial for constructing behaviorally grounded models of pedestrian dynamics. While previous research explored the characteristics of avoidance behaviors, the quantitative description of the dyadic head-on collision avoidance process and decision logic driving these dynamic adjustments remains limited. To address this, we conducted a controlled experiment in which 32 participant pairs performed head-on walking tasks from initial distances of 2, 6, and 10 m. We further developed a novel cost-benefit theoretical framework to model the underlying decision logic. The results demonstrate that head-on collision avoidance is achieved primarily through directional changes, characterized by a distinct pattern in sideways velocity: an initial increase, followed by a decrease and final fluctuations. This process unfolds through four sequential phases: anticipation, action, adjustment, and recovery, exhibiting features of a nonlinear dynamical system. The temporal evolution of turn suddenness and minimum predicted distance provides kinematic evidence for these behavioral transitions. Notably, a cost-benefit model calibrated with 6-m condition effectively predicts the transition from action to adjustment phase under other conditions, demonstrating that the behavioral shift in head-on collision avoidance is governed by a continuous optimization process between safety demands and energy expenditure. This study provides novel insights into the dynamic process and decision-making logic of head-on pedestrian collision avoidance. The proposed framework forms a basis for investigating more complex crowd systems and contributes a behaviorally grounded perspective to the study of interactive pedestrian dynamics in complex systems.