← Back to Sample Lab
SignIQ · curated sampleInteractive viewerPolicy rolloutsReward labels
SignIQ Policy Rollouts — BUDS-class Single-Arm
Single-arm Franka-class (24-D state, 7-D action) · Scripted policy rollouts with reward labels
Episode Summary
InstructionTake the lid off the pot, put the pot on the plate, and use the tool to push to pot to the front of the table.
90
Frames
18.0s
Duration
5 Hz
Control
2
Cameras
24
State dim
7
Action dim
5
Subtasks
success
Outcome
CamerasImage · Wrist Image
Modalitiesvideo × 2 · state (24-D proprioception) · action (7-D)
Format exportsLeRobot · HDF5
LicenseDemo data licensed for inspection; production deliveries are SignIQ commercial.
QA Health
10of 10 checks pass
- Schema parsedPASS
- Timestamps monotonicPASS
- Video ↔ state frame countPASS
- No NaN/Inf in actionsPASS
- Frame-interval jitterPASS
- Outcome label presentPASS
Camera previews
Schema preview
observation.state
shape [24]
- [0]motor_0
- [1]motor_1
- [2]motor_2
- [3]motor_3
- [4]motor_4
- [5]motor_5
- …and 18 more
action
shape [7]
kind:joint_target
control_hz:5 Hz
Interactive inspector
Synchronized cameras, action vs. observed state, joint-space trajectory, and dataset health — driven by the live bundle.
Loading episode bundle…
Why look at this
Policy-rollout episode at 5 Hz with `next.reward` per frame — illustrates the schema we use when capturing rollouts from your deployed policy in our cell, including success/failure labeling and operator-rated severity.
Best for
- →Post-training
- →Evaluation suites
- →Recovery skill learning
- →Reward-aware BC
Capture & license
CaptureSignIQ curated sample
LicenseCommercial — sample for inspection
See full dataset details →