TartanDrive

CMU AirLab · 2022 · datasets.bot · datasets.bot page

One-liner. Large-scale real-world off-road driving dataset (~200k interactions, ~5 hours) collected on a modified Yamaha Viking ATV with seven sensing modalities for learning off-road vehicle dynamics models.

Setup

Schema

Per-timestep multimodal: state/odometry (7D, 50Hz), action/command (2D, 100Hz), RGB image (1024x512, 20Hz), RGB map (501x501, 20Hz), heightmap (501x501, 20Hz), IMU (6D, 200Hz), shock position (4D, 50Hz), wheel RPM (4D, 50Hz), pedals (2D, 50Hz). Maps/heightmaps stereo-derived from MultiSense S21 RGB-D. Stored as rosbags, convertible to torch/numpy.

Links