Datasets:
pick_robosuite v1: "Hand me the {bottle | vial | syringe}" (robosuite, Franka Panda)
Scripted demonstrations of a single-arm Franka Panda picking a named object from a table and placing it in a tray. They were generated in robosuite 1.5.2 (MuJoCo 3.3.7) for the CathLab Companion project (Siemens Healthineers Hackathon 2026, challenge C2).
| Episodes | 384 (128 per object), all expert successes |
| Frames | 69,191 at 20 fps (mean episode length: bottle 216, vial 162, syringe 163 steps) |
| Robot | Franka Panda + Franka hand, robosuite OSC_POSE (delta, base frame) |
| Format | LeRobot v3, robot_type: panda |
| Source | scripted (privileged analytic grasp templates → waypoints) |
I/O (LIBERO-compatible)
| Key | Shape | Meaning |
|---|---|---|
action |
float32[7] | Normalized to [-1, 1]: [dx, dy, dz] / 0.05 m, [drx, dry, drz] / 0.5 rad (axis-angle, robot base frame), gripper (−1 open, +1 close) |
observation.state |
float32[8] | End-effector position (3, base frame, table top at z = 0), end-effector axis-angle (3), 2 finger joint positions |
observation.images.agentview |
256×256×3 video | Fixed third-person camera (robosuite default agentview) |
observation.images.eye_in_hand |
256×256×3 video | Wrist camera |
task |
string | Instruction, one of "Hand me the {object}", "Please pass the {object}", "Can you give me the {object}" |
Images are stored upright (OpenCV convention). LIBERO's raw robosuite frames are vertically flipped, so flip them if you want to match LIBERO preprocessing exactly.
Variation (training grid)
The 384 episodes are the full training grid:
- 3 target objects × 8 layouts × 4 object-colour palettes × 4 table colours;
- per episode, sampled continuously:
- object mass × U(0.7, 1.3);
- friction U(0.5, 1.0);
- light intensity × U(0.7, 1.3);
- colour temperature U(4000, 7500) K.
All three objects are always present (target + 2 distractors). Layouts 8–9, palette 4 and table colour 4 are held out for evaluation and do not appear here.
meta/episodes_extra.jsonl stores, per episode:
- the full
EpisodeConfig(object, layout, colours, physics, lighting, seed); - the instruction and step count;
- the raw shard it came from.
Objects
| Object | Geometry | Mass |
|---|---|---|
| bottle | upright cylinder Ø62 × 250 mm (side grasp) | 0.30 kg |
| vial | upright cylinder Ø27 × 42 mm (top grasp) | 0.02 kg |
| syringe | lying: barrel Ø15 + plunger Ø8 + flange, 140 mm (top grasp across the barrel) | 0.015 kg |
The tray (inner 0.25 × 0.32 m) is fixed to the right of the pick region.
Load
from lerobot.datasets.lerobot_dataset import LeRobotDataset
ds = LeRobotDataset("Rakeshdharan56/pick_robosuite", revision="v1")
Raw per-episode .npz shards (for re-conversion) are in Rakeshdharan56/pick_robosuite_raw.
- Downloads last month
- 76