Analysis of pairwise pedestrian collision avoidance dynamics in T-junctions with different merging setups

Collision avoidance is a critical task in multidirectional pedestrian interactions like merging flows to ensure safety and efficiency. However, the dynamic mechanisms of individual avoidance behavior in different merging situations remain insufficiently understood. This study investigates pairwise pedestrian avoidance maneuvers in T-junctions with several merging flow layouts through controlled experiments, namely head-on and perpendicular encounter avoidance. The avoidance process is discussed in three dimensions: general avoidance outcomes, applied avoidance strategies, and effects of multiple factors on avoidance effectiveness. Results reveal that perpendicular encounter generates longer avoidance distances and time lapses, especially under higher speed levels. Both the moving direction and passing order influence avoidance strategies, while the latter factor induces more pronounced differences in strategic variation (e.g., speed and orientation adjustment). Minimum Predicted Distance (MPD) is introduced as a dynamic metric to quantify the avoidance effectiveness based on mutual adjustment insights, and its variation characteristics indicate disparate avoidance patterns. Specifically, for head-on encounters, pedestrians execute avoidance collaboratively with advance orientation adjustment of leading pedestrians to achieve efficient but collision-free access. For perpendicular encounters, avoidance relies more on speed adjustment of yielding pedestrians due to defensive reactions and certain discrepancies in behavioral tendency, which might affect overall efficiency and balance of merging flow. Our findings underscore the roles of visual conditions and geometric constraints in shaping these nonlinear avoidance dynamics at a T-junction. The work could provide insights for pedestrian dynamics in some merging situations to enhance efficiency and safety levels.