ONTOLOGY VS GRAPH
The ontology defines the language; the graph contains the world.
An ontology can define classes and relations such as Robot, Sensor, Task or Capability. A knowledge graph then instantiates concrete facts: a specific Unitree G1 has a particular model identity, is produced by a particular organization, has evidence from specific sources and may expose particular developer assets.
GRAPH CONTENT
Embodied systems create relationships that flat tables hide.
Identity
Robot, model, configuration and organization identities.
Body
Links, joints, sensors, actuators and other physical components.
Capabilities
What actions or skills the embodiment can support under declared conditions.
Tasks + environment
Which tasks exist and what environmental context they require.
Provenance
Which source, version or observation supports each important assertion.
Lifecycle
How configuration, calibration, capability or evidence changes over time.
EVIDENCE GRAPH
A useful graph should distinguish fact from interpretation.
For OpenRGD, provenance is especially important: imported evidence, editorial summaries, experimental findings and candidate semantics should not collapse into one undifferentiated “truth” record.
OPENRGD.ORG
The website already has the beginnings of this graph — but the site graph is not the standard.
The shared catalog behind OpenRGD.org already separates canonical robot and organization entities from site-specific publication context. That infrastructure can grow into evidence-rich robot pages, but it should remain clearly distinct from the normative OpenRGD specification.