Datasets:
mujoco-franka-reach-spatial
25 MuJoCo Franka reach-and-pre-grasp episodes with ground-truth metric spatial context: an SE(3) frame graph and a queryable relation table (gripper-to-cube distance, signed gripper-to-table clearance, approach speed, time-to-contact) with metric_scale=simulation provenance. Generated with traceplane (pip install traceplane) as a reference for the spatial-context certification dimension: https://traceplane.ai/blog/your-robot-doesnt-know-where-the-table-is. Scene notes: the arm is table-mounted, robot self-collisions are excluded, and arm joints are damped (adjustments to the stock traceplane MuJoCo scene, documented for reproducibility).
QA Status: PASS (0 errors, 0 warnings)
Dataset Summary
| Property | Value |
|---|---|
| Episodes | 25 |
| FPS | 20.0 |
| Robot | franka_panda |
| Format | LeRobot v2.0 |
| Published | 2026-07-27 |
Usage
import traceplane
# Load locally
ds = traceplane.load("mujoco-franka-reach-spatial")
# Or run QA first
report = traceplane.check_dataset("mujoco-franka-reach-spatial")
report.print()
Quality Report
This dataset was validated with traceplane check before publishing.
QA Issues
- [INFO]
META_NO_FEATURES: no 'features' dict in info.json — column schema undocumented - [INFO]
SPATIAL_CONTEXT_SUMMARY: metric spatial QA: 2 relation(s), 25 episode(s), 7000 timestep row(s), scale=simulation.
Published with Traceplane — data infrastructure for robotics trajectories.
- Downloads last month
- 122