The role of torso rotation in pedestrian injury risk during emergency avoidance: A multi-method study integrating...

Pedestrian avoidance behaviors and impact postures significantly influence injury risk and severity in vehicle-pedestrian collisions. However, this relationship remains inadequately understood due to limited access to pedestrian kinematic data in hazardous situations and a lack of empirical research on emergency responses at the moment of impact. This study aimed to quantify how naturalistic avoidance postures affect head and chest injury outcomes compared with normal walking, and to identify the key torso posture parameters that govern injury risk.

Motion‑capture data from a previously published VR‑based experiment, in which volunteers performed emergency avoidance behaviors, were used in this study. The recorded marker trajectories were processed through OpenSim inverse kinematics to reconstruct posture transitions and compute three‑dimensional torso orientation using Euler angles (R1: bending, R2: extension, R3: rotation about the vertical axis). A quantitative staging criterion based on inter‑leg angle changes was applied to distinguish walking from avoidance phases. Extracted postures were then incorporated into MADYMO multibody simulations to evaluate injury outcomes across impact speeds of 20-50 km/h. Head injury was assessed using HIC₁₅, and chest injury was evaluated using the Viscous Criterion (VC) and 3‑ms chest acceleration (C3ms). Statistical analyses included Kernel Density Estimation, Mann‑Whitney U tests, Pearson correlation, random forest feature importance ranking, and linear mixed‑effects models.

Emergency avoidance actions produced complex and rapidly evolving posture trajectories. Compared with normal walking, avoidance postures significantly reduced HIC₁₅ values (all p < 0.05; effect sizes r = 0.36-0.44) across all impact speeds. For chest injuries, avoidance postures increased VC but decreased C3ms relative to walking, indicating a divergent biomechanical response of the thorax. Torso angle distributions differed markedly between walking and avoidance states: R3 (rotation about the vertical axis) emerged as the most discriminative parameter (p = 0.0004; r = 0.52). Under avoidance postures at speeds of 30 km/h and above, R3 showed a strong positive correlation with HIC₁₅ (r > 0.7). Parametric sensitivity analysis further confirmed that R3 exerts a significant independent effect on head injury (β = 9.33, p < 0.001), while its effect on chest injury is divergent, reducing VC but increasing C3ms.

Avoidance postures generally reduce head injury risk compared with normal walking, although this protective effect diminishes as impact speed increases and does not extend consistently to chest injury metrics. Torso rotation about the vertical axis (R3) is the most influential postural parameter governing injury outcomes, correlating strongly with head injury risk at higher speeds and producing divergent effects on chest injury metrics. These findings provide quantitative biomechanical evidence that pedestrian protection systems should account for the diversity of real‑world avoidance configurations rather than relying solely on standardized walking postures. R3 shows potential as a real‑time predictor of head injury risk in pre‑crash scenarios.