Captured site
Harborview Grocery Distribution Annex
Backroom and dock-side grocery distribution layout for tote movement, shelf staging, and pedestrian-interruption checks.
This warehouse robot evaluation page maps available task packs to a Policy Evaluation Run request for one robot profile and validation boundary.
Scenarios
500 ready
Region
Midwest
Evidence
Robot POV available · capture-backed site package
Access
Request-gated
Publication
Publication package complete
Thresholds
Thresholds attached
Episode specs
Episode specs review-required
CPU preflight
CPU setup manifests
GPU handoff
Owner GPU handoff review-required
Site/task simulator run
Run simulator evaluation
Submit a Unitree G1 MuJoCo simulator evaluation request for this captured site. Blueprint queues the request here and forwards it to Pipeline for scheduling.
Scope: simulator only
This request only builds and forwards the simulator evaluation packet for Pipeline execution.
Worker: Blueprint selected
Blueprint chooses the fastest/cheapest available simulator worker for the request.
Robot
Unitree G1
Simulator
MuJoCo
Mode
Queue + forward
WebApp queued
Ready to submit
Pipeline staged
After submit
Provider running
After Pipeline
Complete / failed
When artifacts sync
Artifacts
Pipeline scheduler, worker, runtime, trace, metric, and proof artifacts will appear here after sync. This panel does not create runtime proof by itself.
Proof boundary
WebApp proves request construction, queueing, and forwarding state only. Pipeline owns simulator scheduling, provider execution, runtime logs, and any later pass/fail artifact.
Task packs
Available task packs
These are the tasks a robot team can scope into a Policy Evaluation Run request for this site.
Tote transfer
Tote transfer from dock-side stack to shelf staging
Shelf/bin check
Shelf and bin check with clutter variation
Human crossing
Human crossing route interruption
Included
What is included
The detail page is intentionally compact: enough for a robot team to decide whether this site is worth requesting.
Site card with capture and access posture
Task packs for tote, shelf, and crossing scenarios
Scenario suite summary with route and blocker labels
Task Evaluation Run request path
Capture summary
Walkthrough capture, route notes, task labels, and site package metadata are attached for request review.
Robot POV
Head-height robot POV and aisle-route review are available for scoped evaluation planning.
Geometry/access
Collider and geometry fields are available where applicable, with exact access confirmed per request.
Pipeline status
Manifest-backed request status
These statuses come from synced Pipeline artifact slots or fixtures. They help scope a request before any GPU simulator or owner-system policy run.
Privacy
Rights/privacy packet attached for request review
Retry: Retry after rights ledger or operator approval updates sync.
Failure: Blocked if the rights packet is missing or rejects external use.
World Labs
Compatibility artifacts are support evidence only
Retry: Retry after dependency audit or generated-support asset refresh.
Failure: Dependency failures do not prove or disprove robot readiness.
Materialization
Site, task, scenario, eval, and threshold manifests attached
Retry: Retry materialization sync if card counts or thresholds are missing.
Failure: Partial card families stay Needs review.
CPU preflight
CPU setup manifests present; no owner-system simulator run claimed
Retry: Retry after optional MuJoCo/PyBullet dependency repair.
Failure: CPU failures are setup blockers, not robot policy failures.
GPU handoff
Owner GPU handoff packet available for request-scoped execution
Retry: Retry once owner-system GPU proof artifacts are supplied.
Failure: No GPU handoff packet can be treated as simulator execution.
Eval result
Awaiting policy/container/trace/demo/plugin evidence
Retry: Retry after all six policy modalities have owner-system refs.
Failure: Missing evidence keeps pass/fail and threshold claims blocked.
Policy-improvement export
Policy Improvement Run export awaits request approval
Retry: Retry export after entitlement and package-scope approval.
Failure: Export blockers do not change the site capture provenance.
Evidence boundary
A site page is not a run result.
This page helps a robot team choose a site. A Policy Evaluation Run still needs request review, access confirmation, task scope, robot profile, threshold context, and owner-system proof before any stronger claim is made.