Stage-specific alignment between operational loss and rear-end conflict exposure during traffic oscillations

Stop-and-go oscillations degrade traffic efficiency and elevate rear-end crash risk, yet these two consequences are usually evaluated separately. This study examined the extent to which operational loss and rear-end conflict exposure are aligned within the same traffic oscillation episodes and identified the stage-specific behavioral and kinematic factors associated with their correspondence or divergence.

Operational loss and conflict exposure were quantified using two conceptually parallel time-integrated indicators: Time-Integrated Speed Difference (TISD) and Time-Integrated Time-to-Collision (TIT), respectively. Each oscillation cycle was segmented into pre-deceleration, deceleration, low-speed cruising, and acceleration stages using Wavelet Transform-based transition point detection. Stage-specific behavioral features were estimated with the Asymmetric Behavior (AB) model and summarized as stage-level averages and within-stage adjustment trends. The operational-safety alignment was examined at whole-cycle and stage-specific levels through absolute and platoon-centered comparisons. Feature-block extreme gradient boosting (XGBoost) models and SHapley Additive exPlanations (SHAP) were used to assess the relative contributions of behavioral and kinematic information.

Trajectories of 654 vehicles in 27 oscillation platoons from the CitySim-Freeway B dataset showed a positive but moderate association between TISD and TIT. Whole-cycle correlations were 0.360 for absolute outcomes and 0.335 for platoon-centered outcomes (both p < 0.001). Their stage compositions differed: 72.7% of mean TISD accumulated during low-speed cruising and acceleration, whereas TIT was concentrated in deceleration and low-speed cruising, with deceleration alone accounting for 47.1%. Within-platoon TISD deviations were predominantly explained by behavioral features, while TIT deviations reflected both behavioral and kinematic information. SHAP analysis showed that TISD was mainly associated with pre-deceleration and acceleration-stage behavior, whereas TIT was more strongly linked to behavior and motion states during deceleration, low-speed cruising, and acceleration. Cautious recovery during acceleration was associated with higher operational loss but lower conflict exposure, indicating a potential operational-safety trade-off.

Operational loss and rear-end conflict exposure are related but not interchangeable. Their association varies across oscillation stages and reflects different behavioral and kinematic patterns. The proposed stage-specific framework supports joint evaluation of traffic efficiency and safety and may inform longitudinal control strategies that balance the two objectives.