Prepare an observed replay package
OORB Real2Sim packages connect a recorded robot episode with a URDF or Xacro robot description and replay readiness checks for an Isaac Sim and ROS environment. The current workflow is observed replay of recorded evidence.
Closed-loop policy reruns, predicted task success, and completed replay assembly inside a selected Docker simulator are outside this workflow's current guarantees.
1. Supply the recording and robot description
Prefer an episode so the recording is associated with captured provenance. Raw MCAP inputs are also supported:
runfp real2sim create <episode-id> --urdf robot.urdf --root . --json
Replace the placeholder with an existing episode ID and use your robot's description file. The package records input identities, referenced robot assets, topic mappings, and available local Isaac and ROS evidence. Large MCAPs and assets are referenced rather than copied into the package.
2. Read readiness findings
| Level | Interpretation |
|---|---|
ready | Required input evidence and useful mappings exist, with relevant local replay commands available |
partial | Simulator, ROS, policy, or topic evidence is incomplete |
sensor_only | Sensor evidence can replay, but mapped motion or control evidence is absent |
blocked | Required recording or robot-description input is missing |
Read the detailed checklist as well as the level. Readiness describes the package and inspected environment; it is not a reliability score for the robot's policy.
3. Review a launch plan
Use the returned package ID:
runfp real2sim targets <package-id>
runfp real2sim launch <package-id> --target local --dry-run
The dry run lets you review the commands before launching replay. A local target may offer ROS bag playback when Isaac Sim is unavailable; inspect the target description rather than assuming Isaac will run.
Remote target limits
Linked simulator profiles can supply additional targets. Remote launch depends on accessible input files, target-side commands, and SSH key or agent authentication. When transfer is needed, review and run the supplied staging commands; automatic file synchronization is not provided by this workflow.
Use replay to inspect recorded behavior. For release decisions, combine it with run comparisons and validated task outcomes.