.png){kind=icon}
.png){kind=icon}
Return
Physics-informed graph modeling for traffic shockwave analysis through optimized change-point detection
.png){kind=icon}
PRE
AI
T
X
P
M
B
P
DOI：10.1016/j.trb.2026.103522.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
• A novel two-stage PELT+ framework combines CUSUM-guided candidate pruning with dynamic programming to achieve up to 221 × faster trajectory segmentation than standard PELT while preserving the optimal change-point accuracy. • Shockwave formation and dissipation are reformulated as a physics-informed directed acyclic graph, where edges encode causal braking cascades constrained by theory-grounded propagation speed, temporal, and spatial proximity rules. • The proposed PELT+ + graph method outperforms existing baselines on 120 manually annotated NGSIM STMap segments. • Ablation analysis and sensitivity analyses demonstrate that all design parameters fall within well-behaved linear response regions, confirming the robustness and physical grounding of the framework.
Journal
T
IF:
0
Papers:
78
Citations:
0
Organization
Cited Papers