Litian2002/zerograv-distill-r2-zero-gravity-rendered

Rendered zero-gravity expert trajectories, round-two distillation training set (LeRobot v2.1)

Camera-image version of episodes 0-99 of every task in stray-light/zerograv-manipulation-trajectories-state-v1, rendered so that a vision-language-action model can be trained on them. This is, unchanged, the dataset on which the round-two zero-gravity-trained model of our zero-gravity distillation study was fine-tuned (pi0.5-DROID, openpi LoRA, 42,000 steps, global batch 32). The regular-gravity counterpart, rendered the same way, is Litian2002/zerograv-distill-r2-regular-gravity-rendered.

Contents

800 episodes (8 tasks x 100) and 267,812 frames at 15 fps. One parquet row is one control step.

Task	Prompt	Environment	Source episodes	Episodes	Frames
PushCube	push the cube to the target region	PushCube-DROID-ZeroGrav-OnContact-v1	0-99	100	22,292
PickCube	pick up the cube and hold it at the target location	PickCube-DROID-ZeroGrav-v1	0-99	100	61,246
LiftPeg	grasp the peg and lift it upright	LiftPeg-DROID-ZeroGrav-OnGrasp-WallFix-v1	0-99	100	30,396
PickSingleYCB	pick up the object and hold it at the target location	PickSingleYCB-DROID-ZeroGrav-OnGrasp-v1	0-99	100	49,384
PullCube	Pull the cube to the target position	PullCube-DROID-ZeroGrav-OnContact-Confined-v1	0-99	100	22,281
PlaceSphere	place the sphere into the bin	PlaceSphere-DROID-ZeroGrav-OnGrasp-WallRelease-v2	0-99	100	37,582
RollBall	roll the ball to the goal region	RollBall-DROID-ZeroGrav-OnContact-SpawnBox-Release-Corridor-Still-w3-v1	0-99	100	27,389
PushT	push the T-shaped block onto the T-shaped target	PushT-DROID-ZeroGrav-OnContact-v3	0-99	100	17,242

Feature	Shape	Meaning
exterior_image_1_left	180 x 320 x 3	exterior camera, placed to match the lab's real ZED 2i (intrinsics read from real footage, extrinsics solved against real images)
wrist_image_left	180 x 320 x 3	wrist camera
joint_position	7	arm joint angles, rad
gripper_position	1	0 = open ... 1 = closed
actions	8	[0:7] normalised joint-target increments, 1.0 = 0.2 rad per 15 Hz step (DROID "joint velocity"); [7] gripper command, 0 = open ... 1 = close
timestamp, frame_index, episode_index, index, task_index		LeRobot bookkeeping

The conventions are DROID's, so openpi's DroidInputs reads the data without a remap. Images are stored inside the parquet files (no videos). The layout is LeRobot codebase v2.1, the version that openpi's pinned lerobot reads; it is not the v3.0 layout of the source dataset.

episode_sources.csv has one row per episode: LeRobot episode_index, task, prompt, environment id, source dataset, source episode number, length, whether the goal is visible, and flag (see Filtering).

How it was rendered

• Each recorded simulator state in the source trajectory.h5 was written back into the simulator (no physics stepped) and both cameras were rendered, so robot (Franka + Robotiq 2F-85, the DROID configuration), objects and poses are exactly the expert's. Frame 0 is expert step 0.
• Scene ucr_droid: table and walls textured to look like the lab's real setup, instead of ManiSkill's stock scene. Rasterised (not ray-traced), stored at 320 x 180.
• Goals are visible. In PickCube and PickSingleYCB the goal is a green sphere that the source environments hide; it is drawn here (these two tasks were re-rendered on 2026-10-09 for that).
• The containment walls of the source environments have no visual geometry and do not appear in the images.
• Privileged simulator state (object pose, goal position, contact flags) is not included.

Filtering

No filtering beyond the source dataset's own collection QA: success-only episodes, per-task checks (success on the last step, a 15-frame hold, grasp held at success, arm still after the strike, goal coverage), and trimming to 15 frames after success. See the source card. Episodes 0-99 of each task are used exactly as published. Note that the two source collections used different object-ejection thresholds (1.0 m/s in zero gravity, 4.0 m/s in regular gravity; explained in the regular-gravity source card), and that the source cards' known issues carry over.

One short episode is marked short_unchecked in episode_sources.csv: PushCube source episode 44 (LeRobot episode 44, 24 frames). On video it looks like a success without robot contact (cube already on the target, arm near its start pose), but it was not replay-measured. Every other episode in this set is at least 58 frames long. In the companion regular-gravity set, 33 such episodes were found and replay-verified.

License and attribution

CC BY 4.0, as the source dataset. Expert policies, environments, trajectories and their QA: stray-light/zerograv-manipulation-trajectories-state-v1. Rendering and conversion: distillation/render.py and distillation/to_lerobot.py in the project repository.