AdaFusionNet: Adaptive Trajectory Prediction of Group Two-wheelers in Interactive Conflict Scenarios for Autonomous Vehicles

AdaFusionNet replication package for the article (DOI: 10.26599/JICV.2026.9210102). The package contains STTransformer, RSFM, and GRU fusion code, together with utilities for interaction-scene extraction, recording-level data preparation, sample-window construction, auditing, and prediction alignment. The data directory contains 110 cleaned car and two-wheeler trajectory CSVs. The study uses 45 recording pairs from Tianjin, Changchun, and Xi’an. The additional ChongqingNR files are retained in the source collection but are not used in the study. Install requirements.txt and run ./scripts/run_pipeline.ps1 from the package root. See README.md and REPLICATION_EXPLANATORY_FILE.md for data formats, processing steps, and outputs. Software and documentation are licensed under MIT; trajectory-data rights remain with the data provider.