While advanced driving simulations are being increasingly used in automated driving research, free availability of data and development tools is limited. We present a new open-source framework for synthetic data generation for lane change (LC) intention recognition, targeting prediction of time-to-lane-change in highways. The framework – built on the CARLA driving simulator – advances the state-of-the-art by including a 50-driver dataset, a large-scale 3D map and code for reproducibility and new data generation. The map features a 60 km-long highway track, with a variety of curvature radii and straight road segment lengths. The code covers both simulation enhancements (e.g., dedicated traffic management, car cockpit, engine noise) and baseline Deep Learning model training and evaluation (including CARLA log post-processing to create the input timeseries). The dataset captures over 3,400 annotated LC maneuvers with synchronized ego-vehicle dynamics, road geometry, and surrounding traffic context. Baseline evaluations with state-of-the-art deep learning models confirm its suitability for vehicular applications. The whole framework is released publicly to further support automated driving research.
Sample clips