The SHARC Dataset: A Large Collection of Synthetic Vehicle Trajectory Data on Highway Curves

The SHARC (Synthetic Highway Automotive Road Curves) dataset is a large-scale synthetic vehicle trajectory dataset developed for research in automated driving and deep learning-based trajectory prediction. The dataset was generated using IPG CarMaker, a high-fidelity vehicle and traffic simulation platform, under diverse highway driving scenarios including straight and curved road geometries. It contains multi-vehicle trajectory data from 6,366 lane-change scenarios, designed to address limitations in existing naturalistic driving datasets, particularly the lack of sufficient highway curve scenarios. The dataset provides vehicle state information including position, velocity, acceleration, lane information, vehicle dimensions, vehicle classification, and road-relative coordinates. The data is provided in both CSV and TXT formats to support different research workflows. The SHARC dataset enables reproducible evaluation and development of trajectory prediction models for automated driving applications. The SHARC dataset is accompanied by open-source generation and evaluation tools available at: https://github.com/asha-gamage/SHARC-Dataset