PhaseShift: Topology-Aware Data Harmonization and Model Consolidation Across Signalized Intersections

Source: arXiv cs.RO·2026-08-27 12:00Publish time unverified

arXiv:2608.25275v1 Announce Type: cross Abstract: Learned traffic-behavior models are commonly trained separately for each intersection, creating model portfolios that cannot share evidence across sites. We present PhaseShift, a topology-aware framework that harmonizes heterogeneous roadside trajectories into a shared actor-centric representation and trains one reusable backbone.