Explore the scholarly record.

Abstract

Conventional trajectory-based criticality metrics (e.g., gap and time-to-collision) inherently conflate the interaction demand imposed by a merging vehicle with the evasive or anticipatory response executed by the target follower. Consequently, a low post-event criticality observation cannot distinguish an intrinsically low-demand maneuver from a high-demand intervention absorbed by early driver regulation. Leveraging 25-Hz high-resolution trajectory data from I-24 MOTION, this study establishes a novel demand–regulation–residual decomposition framework. By training a data-driven model-reference state and a conditional contextual gap scale on stable, non-lane-changing interactions, we project interaction demand, anticipatory regulation, and post-response residual gap pressure onto a unified, non-dimensional metric space. Empirical evaluations demonstrate that stronger anticipatory regulation systematically mitigates subsequent residual gap pressure under comparable interaction demand and traffic contexts: across cross-fitted models, prediction RMSE improves by 7.55% across 442 high-continuity events and by 7.17% across a broader domain of 5,803 events. Crucially, by the instant of body-edge intrusion, regulation information is primarily encoded within the follower’s speed–position state. Uncoupling interaction input from driver response shifts trajectory-based safety analysis from static outcome observation to a process-oriented paradigm, offering rigorous benchmarks for microscopic traffic modeling, safety interpretation, and autonomous interaction control.