Answers are grounded in this paper and cite their sources.
This study develops an interaction-conditioned stochastic traffic-state framework that integrates microscopic vehicle interactions with macroscopic traffic variables derived from high-resolution UAV trajectories in heterogeneous, weakly lane-disciplined traffic. Two predominantly near-free-flow trajectory datasets were aggregated into 496 non-overlapping 10-s space–time observations using Edie-consistent area-based density, flow intensity, and mean speed measures. Microscopic same-direction closing interactions were identified using physically constrained leader–follower pairing, and upper-tail inverse time-to-collision and deceleration rate to avoid collision were combined into a standardized Interaction Severity Index (ISI). A hierarchy of deterministic and Gaussian stochastic models was then estimated to determine whether ISI provides additional conditioning information beyond density alone. The density-conditioned heteroskedastic model achieved an RMSE of 30.312 veh/h/m and R 2 of 0.6592, whereas the interaction-conditioned model reduced RMSE to 29.699 veh/h/m and increased R 2 of 0.6728. The likelihood-ratio test supported the additional interaction terms ( p = 0.0009 ) , and AIC favored the interaction-conditioned specification, while BIC favored the more parsimonious density-conditioned model. Blocked temporal cross-validation showed a similarly modest 1.49% reduction in pooled out-of-fold RMSE. The ISI main effect and density–ISI interaction remained statistically detectable after accounting for vehicle-composition diversity, while sensitivity analyses showed generally stable results across plausible temporal-aggregation and interaction-pairing assumptions. Overall, trajectory-derived interaction severity provides incremental state-characterization information within stochastic traffic-flow analysis, although the observed predictive gain is modest and the findings remain specific to the predominantly near-free-flow conditions represented by the analyzed data.
No takes yet. Share an insight, caveat, or question.
Alomari et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: