CAN-Curv: A Vehicle CAN Motion-Control Dataset for Curvature Modeling

The dataset is derived from onboard motion-control and positioning-related signals collected during real-world vehicle driving. After temporal alignment, resampling, segment slicing, local coordinate transformation, and anonymization, the dataset provides structured signals including steering wheel angle, steering rate, vehicle speed, traveled arc length, contextual curvature, yaw rate, and local relative trajectories.

It can support research on CAN-signal-driven curvature modeling, vehicle motion-state estimation, local trajectory reconstruction, and human-like driving behavior learning.To protect data security and privacy, CAN-Curv does not release raw latitude and longitude, absolute timestamps, vehicle identity, driver identity, or specific driving routes. All trajectory information is represented in a segment-local coordinate system with the segment starting point as the origin, and data segments are shuffled to reduce trajectory-stitching risks.