Back to World Models

Light manufacturing

Sonoran Assembly Cart Bay

4025 E University Dr, Phoenix, AZ 85034

A line-side lane with cart staging, station delivery, and return clearance in a tight envelope.

This world model is built from real capture of this facility. Use it to answer a deployment question on the real site, compare the package with hosted evaluation, and decide how your team should test before visiting.

Repeated line-side resupply loops.

Cart fetch and station handoffVision + proprio
SW-PHX-03
cartstationhandoff

Why teams use this site

What teams use this listing for.

The point is to answer one real deployment question before the expensive part starts.

Check deployment fit

See if your robot can move through this site, see the task, and finish the job before a field visit.

Make site-specific data

Render runs from this exact site, vary scenarios, and export outputs for training or debugging.

Compare releases

Run the same task on the same site after each software update so regressions are easy to spot.

Train and demo

Use the exact site for customer demos, operator walkthroughs, and shared remote review.

Site Package

Buy the site package.

Everything your team needs to run its own world model on this facility — walkthrough media, geometry, metadata, and rights.

Walkthrough video, timestamps, and camera poses
Site metadata, workflow notes, and provenance
Geometry / depth artifacts when available from source capture
Package manifest plus rights and privacy terms

Best fit

Teams that want all the site data in their own stack

Starting price

$3,100

This is the path for internal review, integration work, or a team that wants to run or generate its own world model for the site.

Request site package

Hosted evaluation

Request hosted evaluation for this site.

Blueprint runs this site for you. Rerun tasks, review failures, compare checkpoints, and export results — no local setup needed.

Streamed world-model runtime on one exact site
Reset, rerun, and scenario changes
Rollout exports, datasets, and raw bundles
Policy comparison on the same site
Cart At Inbound Lane
Station Requesting Resupply
Clear Return Path

Best fit

Teams that want to run the site now

Self-serve starting rate

$26 / session-hour

A session-hour is one hour of self-serve hosted runtime on one exact site. It covers the live session time used to run, rerun, inspect, and export results. Managed, priority, or higher-touch work is scoped separately when the job needs more support.

Request hosted evaluation

Simple eval loop

What hosted evaluation looks like

One site. One task. One robot question. Start the session, run the task, compare the result, and export what matters.

Step 1

Pick the site

Pick the exact site you want to test so the run stays tied to one real place.

Step 2

Start the hosted session

Blueprint brings up the managed hosted run built from that site.

Step 3

Choose robot, sensors, task, policy/checkpoint, and scenario variation

Choose the robot setup, the task, the policy, and the variation you want to run.

Step 4

Receive the starting observation

The session returns the first view of the site so the policy has a starting point.

Step 5

Let the policy choose an action

The policy decides what to do next, such as move, turn, lift, grasp, or stop.

Step 6

Get the next observation

The hosted world model advances the run one step and returns the next view.

Step 7

Repeat until success, failure, or timeout

Keep looping until the task succeeds, fails, or times out.

Step 8

Score the run, export results, and compare policies

Review the results, export the outputs, and compare checkpoints side by side.

What goes in

  • Exact site selection and workflow lane
  • Autonomous cart tug with front stereo pair
  • Fetch the staged cart and deliver it to the resupply station
  • Clear Lane, Narrow Return Path, Busy Station Handoff

What comes back

  • Starting observation from Sonoran Assembly Cart Bay
  • Step-by-step observations as the run progresses
  • Rollout video, metrics, and failure cases
  • Dataset or raw bundle exports tied to the run
  • Checkpoint comparison on the same site

What teams do with this world model

  • Check deployment fit before travel
  • Generate site-specific data
  • Compare releases on the same setup
  • Train operators or run customer demos

Example

A sample eval loop for Sonoran Assembly Cart Bay

A team picks Sonoran Assembly Cart Bay, chooses Autonomous cart tug with front stereo pair, and tests Resupply checkpoint B on the task to fetch the staged cart and deliver it to the resupply station. They run a few variations like clear lane and narrow return path to see whether the lane is viable, what breaks first, and whether the checkpoint is ready for a real visit. Then they review the rollout video, metrics, failure cases, and exported data.

Related sites