ksha23cc-by-4.0
NeDM Study 4: a Go2 quadruped on CRM granular terrain
A dataset for reproducing Study 4 of 'Learning the Right Abstraction: Neural Reduced Dynamics for Complex Robot Control', containing recorded walking corpora, trained neural surrogate models, fine-tuned policies, and evaluation results for a Unitree Go2 quadruped on granular terrain simulated in Chrono.
Downloads6
Episodes2324
Hours5.7
Why This Matters for Physical AI
This dataset demonstrates how learned neural surrogate models of robot dynamics can improve locomotion control policies on complex granular terrain, advancing practical methods for sim-to-real transfer and efficient policy optimization in legged robotics.
Technical Profile
- Modalities
- proprioceptionimu
- Robot Embodiments
- Unitree Go2quadruped
- Action Space
- joint_positions
- Environment
- simulation
- Task Types
- locomotionwalking
- Episodes
- 2324
- Total Hours
- 5.7
- Data Format
- CSV
- License
- cc-by-4.0
Access
Need custom proprioception data?
Claru builds purpose-built datasets for simulation applications with dense human annotations and quality assurance.
Request a Sample Pack