cawm

Tactile manipulation data for world-model training: real-robot teleoperation on a Franka Panda arm, plus matched simulation rollouts. Most streams are 10 Hz; tool reorientation is 20 Hz and slip-and-catch is 50 Hz.

Preview

Each clip is wrist camera · agentview camera (where one was recorded) · FlexiTac contact map, played at 9.6× real time — the data is 10 Hz, so these are every 8th frame at 12 fps.

These are excerpts, not whole trajectories: a contiguous window of a single episode, chosen automatically as the stretch with the most tactile contact. Contact is sparse (under 10% of pad cells, and many frames have none at all), so a uniformly sampled window would mostly show a black tactile panel. The tactile panel is the merged (24, 32) grid used for training — on the single-pad rigs the upper and lower halves are duplicates of the one physical pad.

Real robot

Franka Panda, teleoperated at 10 Hz.

real/rubik

Rubik's cube manipulation. Wrist + agentview + one FlexiTac pad.
50 episodes · 100,000 steps · 166.7 min · T = 2000

real/allen_insert

Allen-key insertion, single recording session.
80 episodes · 112,792 steps · 188 min · T = 587–2797

real/erase_whiteboard

Whiteboard erasing — sustained sliding contact.
57 episodes · 90,708 steps · 151.2 min · T = 954–2332

real/usb_insertion

USB-connector insertion. See the caveats section below before training.
23 episodes · 62,265 steps · 103.8 min · T = 578–5000

Simulation

sim/slip_catch/hdf5/all + sim/slip_catch/lerobot/all

1,000 trajectories · 96,654 transitions · 50 Hz · variable grip commands. A striped juggling club slips under gravity and is caught by a tactile gripper. Five exploration families vary slip distance, grip timing/strength, pauses and repeated catches. Synchronized wrist/side RGB-D and two tactile pads. All quality-valid outcomes retained. CAWM-compatible HDF5 and LeRobot v3 with original commands and complete transition extras. Whole-trajectory splits: 800 train, 100 validation, 100 test. Dataset documentation · Example

sim/tool_reorient/hdf5/all + sim/tool_reorient/lerobot/all

1,000 trajectories · 290,406 transitions · 20 Hz · fixed closed grip. Palette-knife in-hand reorientation against a support, with synchronized wrist/overview/side cameras and two tactile pads. Five diverse scripted motion families; all quality-valid outcomes retained. CAWM-compatible HDF5 plus LeRobot v3 with original commands. Lossless depth and terminal observations are included for complete world-model transitions. Dataset documentation · Example

sim/stack/hdf5/v2 — corrected Stack Cube v2

1,000 audited successful episodes · 386,544 frames · 40 appearance variants · two RGB-D cameras at 10 Hz. GPT-6.1 Sol Low motion programs, holding each episode's initial end-effector orientation. Small rotational feedback remains; this release replaces the retired hard-coded-downward-target v2.

Dataset documentation · Episode manifest and hashes · Audits · Example video

sim/rubik/hdf5/easy

Cube task, easy randomisation.
500 episodes · 199,500 steps · 332.5 min · T = 399

sim/rubik/hdf5/medium

Cube task, medium randomisation.
500 episodes · 199,500 steps · 332.5 min · T = 399

sim/rubik/hdf5/hard

Cube task, hard randomisation.
900 episodes · 359,100 steps · 598.5 min · T = 399

sim/rubik/hdf5/easy + sim/rubik/lerobot/easy

Two-camera build of the easy split: the agentview stream is recovered from the LeRobot copy and merged onto the HDF5 episodes.
500 episodes · 199,500 steps · 332.5 min · T = 399

sim/pen_in_hole/hdf5/all

Peg-in-hole insertion; variable episode length.
1500 episodes · 389,841 steps · 649.7 min · T = 42–300

Includes the original 1000 episodes plus 500 diverse play episodes added on 2026-10-02. The new episodes cover teacher, hover, surface, rim, insertion/withdrawal, pitch sweep, pitch steps, pitch/gripper coupling, and release; all outcomes are retained. The combined split is 1410 train / 90 validation (new episodes: 470 / 30). The new episodes retain their valid first frame; the original exports are unchanged. All files use the existing wrist-RGB, merged left/right normalized tactile (24×32), and 4D relative-EEF-position/right-gripper action contract. This legacy HDF5 action projection does not include pitch. Source mappings, outcomes, checksums, and exact conversion semantics are in the top-up manifest.

sim/pen_in_hole/lerobot/play_500

Independent, unmerged LeRobot v3 copy of the 500 new play episodes above. 500 episodes · 148,456 frames · 247.4 min · T = 42–300 · 10 Hz

Keeps synchronized front/wrist RGB, separate left/right tactile channels, full robot/object state, and the original 8D XYZ/roll/pitch/yaw/left/right-gripper action commands (roll/yaw zero; pitch and gripper vary). All valid initial frames are retained. Video codec: H.264; per-episode and aggregate statistics are recomputed for this collection. As in the existing LeRobot all copy, all episodes are in its train split; native split labels, profiles, kinematic insertion labels, and termination reasons are recorded in the source manifest. The original sim/pen_in_hole/lerobot/all dataset remains unchanged at 1000 episodes. The two LeRobot folders together contain 1500 unique episodes.

---

real/usb_insertion

23 episodes of a USB-connector insertion task, teleoperated with a 3Dconnexion SpaceMouse and recorded at 10 Hz.

episodes	23
steps	62,265
duration	103.8 min
rate	10 Hz
format	HDF5, one file per episode (episode_0.hdf5 … episode_22.hdf5)

Contents of each file

obs/robot0_eef_pos                      (N, 3)          float32   measured EE position (m)
obs/robot0_eef_quat                     (N, 4)          float32   measured EE orientation, xyzw
obs/robot0_joint_pos                    (N, 7)          float32   joint angles (rad)
obs/gripper_width                       (N, 1)          float32   Franka Hand finger separation (m)
obs/images/cameras_agentview_rgb        (N, 224, 224, 3) uint8    fixed scene camera (RealSense D435i)
obs/images/cameras_wrist_rgb            (N, 224, 224, 3) uint8    wrist camera (RealSense D405)
obs/tactile/tactile_force_grid_left     (N, 16, 32)     float32   FlexiTac contact map, normalised
obs/tactile/tactile_force_grid_right    (N, 16, 32)     float32   see caveat below
actions                                 (N, 7)          float32   see below

File attributes: num_samples, action_dim, init_eef_pos (3,), init_eef_quat (4,).

Action space

7-dimensional, commanded deltas rather than achieved motion:

index	meaning
[0:3]	delta EE position, metres, base frame
[3:6]	delta EE rotation, axis-angle rotvec, radians, base frame (extrinsic)
[6]	gripper: +1 = closed, -1 = open

Observations are captured before the action is applied, so each (obs, action) pair is the state the operator saw and the command they gave in response. The commanded trajectory is init_eef_pos + cumsum(actions[:, :3]); the measured pose is obs/robot0_eef_pos, and the two differ where the impedance controller lagged the target (max 2.4–3.3 cm across these episodes).

Images are stored at the cropped 224×224 resolution used for training. The full uncropped sensor frames were never recorded.

Caveats — please read before training

Several dimensions are constant by construction and carry no signal:

• actions[3:6] is exactly zero in every step of every episode. Orientation was locked during collection, so the arm holds the orientation it homed at for the whole episode. This is a translation-only dataset.
• actions[6] is +1.0 everywhere. The gripper closes before recording starts and never reopens, so no grasp or release event appears in the data.
• gripper_width is ~0 throughout (−0.007 … 0.009 mm). The connector is held by a printed fixture rather than between the fingertips, so the fingers meet.
• tactile_force_grid_right is a byte-exact duplicate of ..._left. The rig has one FlexiTac pad; the second key was added so the schema matches a two-finger setup. It contains no independent measurement.

|dpos| reaches at most 6.00 mm per step (0.06 m/s at 10 Hz).

Provenance and validation

Merged from three collection sessions (2026-08-13 and 2026-08-19); real/usb_insertion/merge_manifest.txt maps each episode back to its source file.

Every episode was checked frame by frame before upload: no duplicate/stalled camera frames, no NaN/inf, consistent dataset lengths, and an identical schema across all 23 files.

License

Not specified by the authors. Please contact the repository owner before use.