Describe the target platform
List the robot configuration, movable components, end effector, and sensors that matter for the intended task. Include the control interface the learner will use. A source with the same broad label, such as a robot arm, may still differ in reach, available observations, or actuation. Begin with the specific compatibility question rather than assuming a category match is sufficient.
Read the source configuration
Ask which hardware and controller versions produced each episode and whether they changed during collection. Identify optional sensors and tools that appear only in some sessions. Record unavailable fields explicitly. A platform name in a catalog does not explain the configuration that generated the delivered observations and actions, particularly when recordings span multiple setups.
Review semantic gaps
Determine whether action fields describe commands, targets, or measured movement. Check coordinate frames, units, and gripper behavior with a representative episode. Matching vector sizes do not establish matching meaning. If adaptation is planned, document the conversion assumptions and the information it requires; some differences may remain unresolved without source details or additional collection.
Test the intended loading path
Load a sample through the buyer’s proposed pipeline and inspect both the data and its interpretation. A video replay can look plausible while action conventions are wrong. Keep technical compatibility, task fit, and proposed permissions as separate review questions. Send Fiund the target platform and required fields so source suitability can be discussed with concrete evidence.
Review checklist
- Document the target configuration
- Identify source hardware versions
- Compare action meaning and units
- Review a sample through the intended loader