A Minimal-Input Framework for Cut-In Detection and Pair-Specific Risk Analysis in Highway Trajectory Data - Traj2Rel-SFC: Trajectory-to-Relation Reconstruction and Context Signatures for Cut-In Interactions

Abstract

Highway cut-ins create new same-lane leader–follower interactions that may require the follower to brake, yet most analysis pipelines depend on dataset-provided lane and neighbour identifiers, limiting reuse across datasets. This thesis presents Traj2Rel SFC, a minimal-input framework that reconstructs lane assignment and same-lane relations from trajectory geometry (x,y) and lane-marking metadata, detects cut-ins as explicit cutter–follower pairs, computes pair-specific surrogate safety measures (DHW, THW, TTC, DRAC), and encodes surrounding traffic context as reversible 16-bit Hilbert space-filling-curve signatures. Evaluated on 60 highD recordings (with highD lane/neighbour identifiers used only as reference labels, and highD indicator columns used only once to calibrate a fixed geometry convention for pairwise SSM computation), the framework achieves mean reconstruction accuracy >0.9998, cut-in detection F1 > 0.999, and context signature agreement of 99.49% over 1.4M stage rows. Multi-indicator analysis shows that THW, TTC, and DRAC capture complementary severity aspects; only 0.08% of events exceed the hard-braking DRAC threshold. Decision-stage features predict execution-stage THW risk with ROC-AUC 0.82 under leave-one-recording-out cross-validation, and SFC context features alone achieve AUC 0.62, showing that spatial context signatures carry standalone predictive signal, although they do not significantly improve AUC over the kinematic-only model. A small exploratory extension on 10 exiD recordings further shows that the same lane-reconstruction and cut-in mining core can be moved to highly interactive highway entry/exit scenes without redesigning the pipeline, although this add-on is intentionally limited and is not presented as a second benchmark. From a software-engineering perspective, the result is a portable and auditable analysis pipeline: method inputs are explicitly restricted, intermediate relations are reconstructable and testable, and outputs can be reproduced without relying on dataset-specific derived fields.