Human representation: consumers and evidence¶

MHR-through-SOMA supplies the nominal body Inkmap places designs on. A nominal body is not a measured robot target: the robot’s compile (tatbot ros compile, ros/README.md section 4.2) takes acquired artwork placed on the printed fiducial page the session measures, and body placement stays out of scope until a measured body-to-surface address mapping exists.

Maintained core¶

Capability

Consumer

Evidence type and limit

Unresolved dependency

Next experiment

Pinned body assets and contracts

Body cache audit (tatbot body bootstrap, tatbot body audit), typed readers (python/tatbot_surface)

Software/cache integrity; no physical geometry acceptance

Reviewed bytes, visual review and device parity

Compare retained poses with the pinned provider

Nominal body and placement

Inkmap browser and its pose showcase

Browser/geometry checks; nominal only

Measured mapping to a physical target

Map one retained placement onto a fixed body-shaped phantom

Artwork and material compilation

tatbot_ink

Deterministic software tests

Measured placement and deposition error

Compare one retained design with marks on the curved silicone skin mount

Release review

Inkmap deployment and public-export gates (config/human-representation/release-gates.json)

Explicit reviewer-bound approvals

Evidence for each independently scoped action

Review the exact candidate revision and its evidence

The ManiSkill simulator carried the rest: the synthetic stroke expert, the registration diagnostics, the evidence reporter, and the frozen proposal-training, phantom-mechanics and anatomy experiments. They were deleted with it on 2026-10-18; their packets are historical. Body placement must not be converted to a printed page’s frame by discarding its address or synthesizing a registration, and offline plans grant no hardware authority.

The next physical result¶

One retained design on the curved silicone skin mount, its printed fiducial page measured, with independent held-out landmarks and observed-versus-planned mark error, scale, shape and uncertainty. Set acceptance limits from task requirements and instrument repeatability before collection. A result on the mount does not qualify the body bridge; that needs a separate body-shaped phantom mapping experiment.